The Boring Stuff That Ruins Trips, Confirmed and Covered
WEEK 97 :: POST 3 :: CLAUDE
Directions Given To The A.I. This Week+
Instructions Given to each A.I. — Please provide 3 prompt variations that share this objective:
Each A.I. also received two static attachments: the blog post template (structure) and the authoring instructions (voice and standards). The text below is the week-specific assignment as sent — reflowed for the web; wording unchanged.
I'd like you to write this week's Ketelsen.ai post. Two files are attached: the blog post template (the structure to follow) and the authoring instructions (context, voice, and standards). Please read both before you begin, then produce the complete post in a single response.
This week's theme: "The Boring Stuff That Ruins Trips" — Logistics, Risk and Protection.
This is Week 6 of an eight-week series on planning a vacation with AI. The reader now has the shape of their trip: a destination, booked flights and lodging, and a day-by-day plan. This week handles the unglamorous details that carry the biggest downside if they go wrong — the travel equivalent of the finance-and-insurance desk, where nobody enjoys the paperwork and a missed line can cost the whole trip.
The topics are the ones travellers skip until it is too late: passports, visas, and entry requirements; travel insurance and what it actually covers versus the checkout-box upsell; health preparation such as vaccinations and carrying medication across borders; a phone and data strategy; a payment strategy that accounts for foreign-transaction fees and local cash-versus-card norms; and a document-and-backup protocol for when a wallet or phone goes missing. Each is low-glamour and high-consequence, which is exactly why a structured pass with AI is worth the reader's time.
The deliverable the reader should walk away holding is a personalised risk-and-requirements audit for their specific destination and their own traveller profile, turned into a countdown checklist — tasks grouped by how many days before departure they must be done, from the far-out items down to the final week.
The three prompts should help a reader work through:
- Requirements and documents, sorted by deadline. What this traveller, on this passport, going to this destination, needs to arrange and by when — and crucially, where the official answer lives, because this is the one area where an out-of-date answer can mean being turned around at the border.
- Insurance and money, read honestly. Turning a policy or a payment setup into plain language: what a given travel-insurance tier really covers versus the upsell, and how foreign-transaction fees and cash norms change what the reader should carry and which card they should use. The AI is good at explaining categories and the questions to ask; the reader supplies the actual policy and card terms.
- A protection and backup protocol. A simple, personal system for documents, medication, and emergency contacts — what to copy, where to store it, and what to do first if a phone or wallet disappears mid-trip.
At the advanced tier, the strongest version of this week is a countdown audit matrix: requirement or task as rows; the responsible source, the deadline window, and the reader's current status as columns; sorted so the earliest deadlines surface first. That structure is worth reaching for.
A hard constraint, and it matters more here than anywhere in the series. AI models cannot see current visa rules, this year's entry requirements, a specific insurance policy's fine print, or today's vaccination guidance — and these are jurisdiction-specific, change without notice, and are exactly the questions where a confident wrong answer is dangerous. No prompt in this post may ask the AI to state current entry or visa requirements as fact, confirm what a named insurance policy covers, or give definitive medical or legal guidance. Prompts must have the AI produce what to check and which official source to check it against — the government page, the insurer, the pharmacy, the consulate — rather than deliver a ruling the reader might act on without verifying. Say this plainly inside the prompts themselves.
Design the prompts so the AI does what it is genuinely good at: turning a vague sense of "I should probably sort out insurance" into a specific, sequenced list of what to confirm, with whom, and by when. The reader supplies their profile and their destination; the AI supplies the structured audit and points them at authoritative sources. Posts that have the AI assert current requirements as settled fact should expect to be marked down on Practical Utility and Content Accuracy.
Series dependency chain, for the Metadata block: Week 6 consumes the destination from Week 2 and the booked flights and lodging from Weeks 3 and 4 (the countdown is anchored to the confirmed departure date, and entry requirements depend on the specific destination). Week 6 produces the requirements audit and the protection protocol, which Week 7's in-trip prompts lean on when something goes wrong on the ground and Week 8's reconciliation reuses when closing out claims and disputes.
Because readers may arrive at this post without having read Weeks 1 to 5, the prompts should work for someone who knows their destination and rough travel dates, while making clear they get far more from them with confirmed bookings and a real traveller profile in hand.
Three difficulty tiers as always — Beginner, Intermediate, Advanced — each a genuinely different approach to the same problem, not the same prompt at three lengths.
On examples: this is a consumer travel topic. The template lists tech startup / retail / freelance as suggested industry examples — those are marked MAY, and this week you should almost certainly adapt them. A family sorting children's passports and a parent's medication supply, a couple comparing a card's foreign-transaction fees, a solo traveller building an emergency-contact and document-backup kit, and an older traveller coordinating prescriptions across a long trip are the right contexts here. Choosing them over the suggested business examples is correct behaviour and will not be scored against you.
A note on supplied figures. Anything marked `[SUPPLIED — use as given]` above came from Ketelsen.ai's own research brief. Use it freely — you are not fabricating by repeating it, and you will not be marked down for leaving it uncited. Do not attach an invented source to it. (No supplied figures this week. Given the requirements-change-without-notice constraint above, this is a bad week to invent any — if you find yourself reaching for a specific visa fee, a coverage limit, or a vaccination requirement, that is the signal to restructure the prompt so the reader confirms the real figure at the source.)
## BEFORE YOU SUBMIT — STRUCTURAL CHECK
(This block is identical every week. It exists because these specific items are the ones posts drop, and a dropped structural item costs compliance points for something that takes one minute to add.)
Your post is parsed by a script before any human reads it. Confirm all seven:
1. ☐ Response begins with `PLATFORM: <your name>` and `WEEK: 6` 2. ☐ `## Lead` present once, at the very top, before Variation 1 3. ☐ `## In one line` present in all three variations 4. ☐ `## What this prompt gives you` present in all three variations 5. ☐ `## The Prompt` present in all three variations, with the prompt in double quotes beneath it 6. ☐ `## Introductory Hook` and `## Current Use` present in all three variations (three of each — not one) 7. ☐ Every template heading written as `##`, none bolded instead; prompt breakdown is running text split on ` : `, with no `###` headings inside it
A complete post has 57 `##` headings. If your count is well short, a section is missing or was bolded instead of hashed.
One extra check this week: confirm no prompt asks the AI to state current visa or entry requirements as fact, confirm a named insurance policy's coverage, or give definitive medical or legal guidance. Those must be things the reader verifies at an official source.
The parts of a trip that actually ruin trips are never the parts you daydreamed about — it's the passport with four months of validity left, the insurance tier you clicked past at checkout, the card quietly taking three percent of every dinner. This week's three prompts turn that low hum of unease into a sequenced, sourced audit: a beginner checklist that tells you what to confirm and who can confirm it, an intermediate pass that reads your own policy and card terms back to you in plain English, and an advanced countdown matrix that dates every task backwards from departure and tells you exactly what to do the moment a wallet disappears. None of them will tell you what the rules are. That restraint is the entire reason they're safe to use, and it's what makes them faster than guessing.
The Pre-Departure Reality Check
Turn vague travel dread into a dated, sourced to-do list.
You have the destination. You have flights, you have somewhere to sleep, you have a rough plan for the days. And you still have a faint, unplaceable feeling that something is going to go wrong at an airport counter. That feeling is usually correct, and it is almost never about the interesting parts of the trip. It's about the fact that nobody ever hands you the list — the boring inventory of documents, appointments, phone calls and copies that stands between a booked trip and a trip that actually happens. Most travellers assemble that list from memory, panic and a friend's anecdote, roughly six days before departure. This prompt hands you the list on day one, in the order the tasks have to be done.
Why this matters now
Lead time is the thing that bites, and it bites earliest. A renewed passport, an appointment at a travel health clinic, a visa application, a replacement bank card sent to your home address — every one of those is measured in weeks, and every one of them is invisible until you go looking. Running this prompt the week you book, rather than the week you fly, is worth more than any other single hour you'll spend on the trip. It's also the moment when the answers are cheapest to act on, because you still have room to be wrong about one of them.
I'm planning a trip and I want to make sure I haven't missed anything boring but important. Here are my details.
Destinations and any countries I'll transit through: [COUNTRIES]
Passport or passports I hold, and the issuing country: [PASSPORT DETAILS]
Departure date and return date: [DATES]
Who is travelling: [FOR EXAMPLE, TWO ADULTS AND A SEVEN-YEAR-OLD]
Anything unusual about us or this trip: [FOR EXAMPLE, PRESCRIPTION MEDICATION, A CHILD'S FIRST PASSPORT, A RENTAL CAR, A WORK LAPTOP, A LONG STAY, TRAVELLING WITH A PET]
Important instruction, please follow it exactly. Do not tell me what the current visa, entry, health or insurance rules actually are. Those rules change without notice, they differ by nationality, and a confident wrong answer here is worse than no answer. I will verify every single one of them myself. Your job is to tell me what to check and who the authoritative source is, not what the answer will be.
Give me a checklist of everything I need to sort out before this trip. Cover at minimum: passports and entry documents, travel insurance, health and medication, phone and data, money and payments, and copies or backups of important documents.
For each item on the list, give me three things on one line. First, what I need to confirm, phrased as a question I can answer yes or no. Second, the type of official source that can answer it — for example my own government's passport agency, the destination country's embassy or consulate, my airline, my insurer, my bank, my pharmacist or a travel health clinic. Third, roughly how far ahead of departure this normally needs to be started.
Sort the whole list so the items that need the longest lead time appear first.
At the end, flag the two or three items that would be the most expensive or the most trip-ending if I got them wrong, and explain why those ones in particular.
Finally, if anything in my details makes a specific item unusually likely to bite me, say so plainly.
How the AI reads this prompt
Practical examples from different industries
A family of four with a summer trip booked. The parents fill in two adult passports, one child passport expiring in eight months, and one seven-year-old who has never had a passport, plus a note that one parent takes a daily prescription. The output opens with the child's first passport application — the longest lead time on the list, and the one that also depends on gathering a birth certificate and both parents' consent — then works down through the second child's passport validity, a pharmacy conversation about a supply large enough to cover the trip plus a delay, and only then reaches insurance and phone plans. The parents' own mental list had started with insurance, which is why the sort order matters.
A solo traveller on a two-week trip with a connection through a third country. The unusual field says one prescription medication and a laptop for remote work. The checklist surfaces the transit country as its own line with its own source — the transit country's consulate rather than the destination's — flags carrying medication in original labelled packaging with a doctor's letter as something to ask a pharmacist about rather than something the AI rules on, and puts the phone and data strategy earlier than expected because a work trip that loses connectivity loses income. The traveller had not considered that a layover could have entry requirements at all.
A retired couple planning a five-week road trip that crosses an international border by car. Their unusual field mentions a rental car, two sets of daily medications and a five-week duration. The list separates vehicle documentation and cross-border insurance from personal insurance, notes that a long stay can change which permissions apply and points them at the destination's consulate to confirm, and raises the question of whether a five-week medication supply exceeds what their pharmacy dispenses in one go — a question with a four-week lead time that would otherwise be discovered in week four of the trip.
Creative use case ideas
Reframe it for a house move. Swap destination for new address and departure date for closing date, and the same structure produces a deadline-sorted list of utilities, address changes, school records, insurance transfers and the source who can confirm each. Moves fail for exactly the same reason trips do — a long-lead item nobody knew was on the list.
Use it for a student heading abroad for a semester. Enrolment paperwork, student visa categories, health cover that works with a university's requirements and a bank account that survives a year overseas all have lead times measured in months, and a nineteen-year-old doing this for the first time has no prior list to work from.
Run it for a community group or sports club travelling together. The unusual field carries the group-specific complications — a minor travelling without a parent, shared equipment crossing a border, a group insurance question — and the output becomes something an organiser can hand to fifteen families with individual items marked.
Point it at a wedding abroad. Marriage documentation requirements, residency periods, document legalisation and translation all sit exactly in this prompt's sweet spot, and all are things where the AI naming the right office to call is worth more than any answer it could invent.
Adapt it for a freelancer's first international client visit. Work devices, equipment, what you can and cannot say about the purpose of the trip at a border, and whether business activity changes the category of permission needed are all questions that go badly when improvised at a counter, and all of them resolve to a source rather than an answer.
Adaptability tips
The fastest useful change is to the six domains. If your trip has a dimension the standard list doesn't cover — a pet, a vehicle, sports equipment, a musical instrument, professional gear that a customs officer might read as commercial — add it as a seventh named domain and the model will build it a full branch instead of squeezing it into an existing one.
Change the sort key and you change what the list is for. Sorting by lead time gives you a schedule. Sorting by who has to do it gives you something you can split across two people. Sorting by cost of failure gives you a list to work through when you only have one evening left before you fly.
Scale it down for a short domestic trip by cutting the entry-documents domain entirely and keeping the rest. Scale it up for a multi-country trip by running it once per country and asking, in a second message, which items are shared and which are country-specific — that second pass is where the duplication collapses.
If you are running it for someone else — an ageing parent, a young adult, a friend who is worse at this than you are — write their details into the fields and add a line asking for the output to be phrased as instructions to them rather than notes to you. Same list, entirely different usability.
Pro tips
Run it twice, once with your inputs as written and once with a deliberately paranoid version of the unusual field where you list every complication you can think of even if you're not sure it applies. The second run will surface items the first missed, and discarding an item that doesn't apply takes five seconds.
When the output lands, immediately paste it back with a single instruction — convert every line into a calendar entry with a date, calculated backwards from my departure date of [DATE]. The model handles the arithmetic and you get something you can actually put in a calendar rather than something you have to transcribe.
Ask for the list twice at different specificity levels. A one-line version per domain is what you skim on a phone; the full version is what you work through at a desk. The same content at two resolutions is more useful than either alone.
Keep the completed output. Ninety percent of it is reusable for your next trip, and the ten percent that changes is the ten percent you should have been re-checking anyway.
Prerequisites
Know where you are going, including any connections or layovers, and have at least approximate travel dates. Know which passport each traveller will use and which country issued it. Have your passports physically in front of you, or at least the expiry dates, because passport validity is the first question the output will ask and the one most likely to change what you do this week. Nothing else is required — this prompt is designed to work before any bookings are confirmed.
Required tools
Any general-purpose conversational AI tool — ChatGPT, Claude, Gemini or an equivalent — on any tier, including free tiers. No web browsing, no plugins and no file uploads are needed, and browsing is not a substitute for verification here in any case. A calendar app is useful for the follow-up step but is not required.
Frequently asked questions
Why won't the prompt just tell me whether I need a visa?
Because a language model has no live view of any country's entry rules, and those rules are nationality-specific and change without announcement. The model can produce a confident, well-written, wrong answer to that question with no warning signal attached, and the cost of acting on it is being denied boarding or turned around at a border. What it can do reliably is tell you that the question exists for your specific combination of passport and destination, and tell you whose page carries the authoritative answer. That's a genuinely useful division of labour.
How far ahead should I actually run this?
The moment you have a destination and rough dates, which is usually before you book anything. The most valuable items on the list are the ones with lead times of a month or more, and those are exactly the ones that become impossible rather than merely stressful if you find them late. Running it a second time about three weeks out, with your statuses filled in, catches anything that has drifted.
The output gave me a lead time that seems wrong.
Treat every lead time as a rough planning figure rather than a published processing time, and confirm the real one with the office in question — which is precisely what the source column is there for. A lead-time estimate that's a week optimistic still gets you started in time, which is why the prompt is allowed to give you one at all. If an estimate seems wildly off for your country, it probably is, and the fix is a two-minute check rather than a better prompt.
Can I skip this if I've travelled to the destination before?
The list is shorter for a repeat trip, but it is not empty, and the items that change between visits are the ones with consequences. Passport validity moves on its own. Card terms and insurance policies renew with different wording. Entry requirements are the category most likely to have changed since your last visit and the least likely to announce it. Run it, then discard the two-thirds you've already handled.
What if the model refuses to answer part of this?
It shouldn't, because the prompt never asks it to make a call it isn't equipped to make. If it does hedge — usually by adding disclaimers to every line — add a sentence saying you understand it cannot confirm current requirements and you are only asking for the checklist and the source types. The hedging is a sign the model thinks you might act on its output as fact, and reassuring it on that point clears it.
Recommended follow-up prompts
The Week 2 destination-selection prompt from this series, if you have not settled a destination yet — the checklist gets far more precise once the country is fixed, and running these in the wrong order wastes both.
A calendar-conversion follow-up, run as a second message on the same conversation: ask for every checklist item to be converted into a dated task counting backwards from your departure date, grouped by month.
A delegation prompt, for travelling parties of more than one: ask for the completed checklist to be split into two lists by who is best placed to do each item, with anything requiring both people flagged separately.
Tags and categories
Tags:
travel planning, travel checklist, passports, travel insurance, travel documents, pre-trip preparation, deadlines, source verification, beginner prompts, risk management
Categories:
Travel & Trip Planning, Personal Productivity
Citations
U.S. Department of State, Bureau of Consular Affairs — travel.state.gov. The authoritative source for U.S. passport services and country-specific information for U.S. travellers.
Centers for Disease Control and Prevention, Travelers' Health — wwwnc.cdc.gov/travel. Destination-specific health guidance for U.S. travellers, referenced here as an example of the source type the prompt directs readers toward.
UK Foreign, Commonwealth & Development Office foreign travel advice — gov.uk/foreign-travel-advice. The equivalent authoritative source for UK travellers, included to illustrate that the correct source depends on the traveller's nationality rather than the destination.
IATA Travel Centre — iatatravelcentre.com. An airline-industry reference for passport, visa and health documentation requirements, useful as a cross-check but not a substitute for the issuing government or destination consulate.
The Fine Print Translator
Read your own policy and card terms in plain English.
You almost certainly own travel insurance you have never read. It came bundled with a card, or you clicked the recommended tier at checkout because the alternative was a page of comparison, and ever since it has sat in your inbox radiating a vague sense of being handled. It is probably not handled. The gap between what people think their policy covers and what the wording says is where the genuinely expensive travel stories come from — the cancelled trip that wasn't cancelled for a covered reason, the stolen camera that exceeded the single-item limit, the claim rejected for a receipt nobody thought to keep. This prompt does not tell you what you're covered for. It does something more useful and more honest: it reads your own document back to you in language you can act on, and hands you the questions that make a call to your insurer take four minutes instead of forty.
Why this matters now
The information asymmetry here is not accidental — policy wording and card terms are written to be technically complete rather than legible, and the reader who most needs them is the reader least equipped to parse them. What has changed recently is that you can now paste that document into a tool that will genuinely translate it without charging you or selling you anything. The window to use this is now, before departure, because almost every meaningful lever in a travel insurance policy or a card's fee structure can only be pulled in advance. After you fly, you are reading your policy to find out what happened to you rather than to decide what to do.
Act as a careful, plain-spoken explainer who helps people understand documents they have paid for but never read. You are not an insurance broker, a lawyer or a financial adviser, and you must not tell me whether I am covered for anything. Your job is to help me read what is in front of me and work out what to ask.
Below I will paste material of my own. It may be the summary page of a travel insurance policy, the terms for a credit or debit card, or both.
[PASTE YOUR POLICY SUMMARY OR CARD TERMS HERE]
My trip and my situation. Destination: [PLACE]. Dates: [DATES]. Who is travelling: [PEOPLE AND AGES]. Things that might matter: [FOR EXAMPLE A PRE-EXISTING CONDITION, EXPENSIVE EQUIPMENT, A RENTAL CAR, AN ADVENTURE ACTIVITY, A LONG STAY, A NON-REFUNDABLE BOOKING].
Work through it in three passes and label each one clearly.
Pass 1, translate. Restate what this document appears to say in plain English, section by section, in the order the document presents it. Where the wording is ambiguous, or does more than one thing at once, say that it is ambiguous rather than resolving it for me.
Pass 2, find the edges. Point out where cover typically stops or costs typically change in documents of this type. For insurance, look for excesses and deductibles, per-item and per-category limits, exclusions, activity and age restrictions, pre-existing condition wording, notification deadlines, and what evidence a claim requires. For a card, look for foreign transaction fees, ATM and cash advance fees, dynamic currency conversion, holds placed by hotels and rental companies, and any charge that only appears abroad. For each edge you flag, quote the exact words in my document that made you flag it. If my document does not address something on that list at all, say so explicitly. Silence in the document is the finding I most need from you.
Pass 3, give me the call list. Write the questions I should put to the insurer or the bank, in the order I should ask them, phrased so that a call centre agent can answer each one without hedging. Mark the three questions that matter most for my specific trip and situation. For each of those three, tell me what a clear answer sounds like and what an evasive answer sounds like.
Rules for the whole task. Do not conclude that I am or am not covered for anything, and do not estimate what a claim would pay out. If I ask you to, decline and give me the question to ask instead. Do not fill gaps in my document with what policies usually say — mark the gap and move on.
How the AI reads this prompt
Practical examples from different industries
A couple comparing two credit cards before a three-week trip. They paste the fee schedules for both, note that they expect to spend heavily on restaurants and to rent a car. Pass one translates each fee schedule into plain terms. Pass two flags the foreign transaction percentage on one card, the ATM withdrawal structure on the other, the hold a rental company will place against the available balance, and — critically — the fact that neither document mentions dynamic currency conversion, which is where the largest avoidable losses usually occur. Pass three gives them four questions for each bank. They pick a card in an evening rather than guessing.
A traveller with a managed pre-existing condition reading a policy summary. They paste the summary and disclose the condition in the situation field. The model quotes back the exact pre-existing condition wording, notes where the summary refers out to a full policy document it has not been given, and flags that the medical declaration process and its deadline are not described in what was pasted. Pass three produces a call list led by the one question that decides everything — whether the condition needs to be declared and accepted before departure for medical cover to apply at all. The prompt never states the answer, which is right, because only the insurer can give it.
A family relying on the travel insurance bundled with a credit card. They paste the card's benefits guide and describe two adults, two children and a non-refundable package holiday. Pass two surfaces per-person versus per-trip limits, whether cover depends on paying for the trip with that specific card, and how children are treated when they are not named on the account. Pass three gives them a call list for the card issuer's benefits administrator rather than the bank's general line, which is a distinction most people discover only when they try to claim.
Creative use case ideas
Run it on a rental lease or tenancy agreement before you sign. Deposit conditions, notice periods, what counts as damage and what the landlord is obliged to do are exactly the sort of clauses that read as boilerplate and behave as landmines, and the call list becomes the questions to ask before you commit rather than after.
Point it at a phone or broadband contract, particularly the international roaming terms — which is a direct extension of this week's topic and answers the question of whether you need a local eSIM at all or already have something adequate.
Use it on a pet insurance policy, a home contents policy, or the one that catches most people out, the section of a home policy dealing with belongings taken outside the home. Travellers routinely buy separate cover for a laptop that is already covered, or assume cover that stops at the front door.
Try it on an employer benefits enrolment pack. Most of these contain a travel or medical component nobody reads, and the discovery that your employer already provides emergency medical cover abroad has real financial consequences for what you buy.
Apply it to a gym membership, a subscription service or a course enrolment before cancelling anything. Notice periods, automatic renewal wording and refund conditions all follow the same document pattern, and the same three passes work unchanged.
Adaptability tips
Swap the domain-specific list in pass two and the prompt retrains itself on any document type. For a rental agreement, replace excesses and exclusions with notice periods, deposit conditions and repair obligations. The three-pass skeleton is the reusable part.
If your document is long, paste it in sections and run pass one on each, then ask for pass two across all of them together. Models handle a long document better when the reading is separated from the analysis, and you will get noticeably fewer skipped clauses.
Add a comparison dimension when you have two documents. Give both, and add an instruction asking where they differ in ways that matter for your specific trip, with a note that it should not recommend one. The comparison is genuinely useful; the recommendation would be exactly the kind of ruling this prompt exists to avoid.
For a reader who will make the phone call themselves and does not need the reasoning, ask for pass three only, formatted as a numbered script with space to write the answers. Same prompt, cut down to the part that gets used.
Pro tips
After pass three, ask one more question in the same conversation — what would you need to see from me to answer the three flagged questions yourself? The answer tells you exactly which document you are missing, and it is usually the full policy wording rather than the summary you pasted.
Record the answers your insurer or bank gives you back into the same conversation, then ask for the practical implications for how you pack, what you photograph, and what receipts you keep. The value of a policy call is mostly in the behaviour it changes before you leave.
If the model starts hedging every sentence, it usually means the pasted material is too thin to support pass two. That hedging is a signal rather than a failure — go and find the full document.
Keep the call list and the answers together in one file. When something goes wrong on the trip, the notification deadline and the evidence requirement are the two things you will need within hours, and they are the two things nobody can find.
Prerequisites
You need the actual documents. A policy summary or certificate, a card's terms and fee schedule, or a benefits guide — the real text, not your memory of it. Screenshots will not paste usefully into most tools, so find the text version, which is usually available in your account portal or the original confirmation email. You also need a clear description of your own situation, including anything you might be tempted to leave out, since the omitted detail is disproportionately likely to be the one that matters. Basic comfort with pasting a long block of text into an AI tool is assumed.
Required tools
A general-purpose AI tool that accepts long pasted text — ChatGPT, Claude or Gemini all work, including on free tiers, though a longer context window helps with full policy wordings rather than summaries. File upload is convenient if your policy is a PDF but is not required, since copied text works identically. No browsing or plugins are needed, and a browsing tool actively works against this prompt by tempting the model to look up what policies generally say instead of reading yours.
Frequently asked questions
Is it safe to paste my policy or card terms into an AI tool?
Policy wordings and fee schedules are standard documents rather than personal data, so the wording itself is low risk. What you should strip before pasting is the personal layer — policy numbers, account numbers, card numbers, dates of birth and addresses, none of which the prompt needs to do its job. Check your tool's data retention settings if the material is sensitive, and consider whether your workplace has a policy about it. The analysis works just as well on a redacted document.
Why won't it just tell me whether I'm covered?
Because whether you are covered depends on the full policy wording, endorsements, your declarations, and how the insurer applies all of that to your specific circumstances — none of which is fully present in what you pasted. An answer that sounds authoritative and turns out to be wrong is worse than no answer, because it stops you making the phone call that would have caught it. The prompt is built to make that call short and productive instead of replacing it.
The output quoted something that isn't in my document.
That is the one failure mode to watch for, and the quote requirement in pass two exists to make it visible. Search your document for the quoted phrase, and if it is not there, say so in the conversation and ask for pass two to be redone using only text present in what you pasted. Models occasionally blend general knowledge into document analysis, and calling it out in the conversation corrects it reliably.
Can I use this after something has already gone wrong on my trip?
Yes, and it is worth doing, but the emphasis shifts. Ask for pass two to focus on notification deadlines and evidence requirements first, since those are the parts that expire, and treat pass three as questions for a claims handler rather than a sales line. The one thing to do before any of it is report the incident to your insurer within whatever window your policy specifies, because that window closes whether or not you have finished reading.
My policy is fifty pages. Does this still work?
It works better, but you have to feed it properly. Paste it in sections and run the translation pass on each section before asking for the edge analysis across the whole thing. The alternative — asking for everything at once on a very long document — produces analysis that is real for the first few pages and thin for the rest, which is the least useful kind of output because the thinness is invisible.
Recommended follow-up prompts
A claims-readiness prompt, run as a second message: ask for a short list of what to photograph, keep and record during the trip so that any claim your policy does support can actually be evidenced.
The Week 4 lodging prompt from this series, revisited with your cancellation terms pasted in — the same three-pass structure applied to a booking confirmation surfaces refund conditions people routinely misremember.
A currency and cash-strategy prompt: having decided which card to use, ask for a plan covering how much local cash to carry, when cards are typically not accepted at your destination, and what to do about the card holds a rental company or hotel may place.
Tags and categories
Tags:
travel insurance, credit card fees, foreign transaction fees, policy wording, document analysis, plain language, claims preparation, intermediate prompts, question generation, consumer finance
Categories:
Travel & Trip Planning, Personal Finance
Citations
Consumer Financial Protection Bureau — consumerfinance.gov. U.S. federal consumer resource on credit card terms, fees and disclosures, cited here as an example of an authoritative non-commercial source on the card side of this prompt.
U.S. Department of State, Bureau of Consular Affairs — travel.state.gov. Carries the department's guidance for travellers on insurance and medical cover abroad, and is the source type the prompt's call list is intended to complement rather than replace.
UK government travel insurance guidance — gov.uk. The UK equivalent consumer guidance, included because insurance norms, terminology and regulatory protections differ by country and the reader's own jurisdiction determines which guidance applies.
The Countdown Audit Matrix
Build a re-runnable countdown that surfaces the earliest deadline first.
A checklist tells you what to do. It does not tell you what cannot start until something else finishes, which is the failure that actually strands people. The visa application that needs a passport with more validity than yours currently has. The medication letter that needs a prescription renewal that needs an appointment. The cancellation cover that only exists from the moment you buy the policy, meaning every day you delay is a day of the trip you cannot protect retroactively. These are dependency chains, and a flat list hides them completely — you tick off nine easy items and discover the tenth needed to be started three weeks ago. This prompt builds the version that accounts for them, dates every task backwards from your departure, criticises its own output, and comes back to life next week when you paste your updated statuses into it.
Why this matters now
The reason to build this now rather than closer to departure is that its entire value is in the long-lead items, and those stop being actionable in a way that no amount of urgency later can fix. It also matters that this version is designed to be re-run. A trip plan is not a document you write once — it is a state that changes as bookings confirm, applications return and requirements get verified, and most planning tools quietly assume otherwise. Building something you will update three or four times before you leave is the difference between a plan you trust and a plan you re-derive from anxiety every Sunday evening.
You are going to help me build a countdown audit for a trip. The output is a working document I will update and re-run until I leave, not a one-off answer. Work through the numbered stages in order and do not skip ahead.
My inputs.
Today's date: [DATE]. Departure date: [DATE]. Return date: [DATE].
Destination and any transit points: [PLACES].
Travellers, with passport issuing country and age: [LIST].
Bookings confirmed so far: [FLIGHTS, LODGING, CAR, ACTIVITIES].
Constraints and complications: [FOR EXAMPLE PRESCRIPTION MEDICATION, A PRE-EXISTING CONDITION, MOBILITY NEEDS, A CHILD'S FIRST PASSPORT, DUAL NATIONALITY, A WORK DEVICE, A PET, EXPENSIVE EQUIPMENT, A RENTAL CAR, A STAY LONGER THAN A MONTH].
Standing rule for the entire task. You do not know current visa rules, entry requirements, vaccination guidance, or the terms of any specific insurance policy or card, and you must never state one as fact. Every row you produce must end in a verification step I perform against a named type of authoritative source. If you find yourself about to assert a requirement, convert it into a question and a source instead. Apply this rule to every stage below without me repeating it.
Stage 1, inventory. From my inputs, derive every requirement, task and decision that could plausibly apply, across six domains — entry documents, insurance, health and medication, connectivity, money, and document backup. Include items that may not apply to me and mark those as confirm whether this applies. Over-inclusion is cheap at this stage. Omission is not.
Stage 2, build the matrix. Output one row per item with these fields, separated by a middle dot. Item. Question I need answered. Type of authoritative source. Typical lead time. Deadline date, calculated backwards from my departure date. Consequence if missed, rated one to five where five means the trip does not happen. Depends on, naming another item or none. Status, set to not started for everything. Sort by deadline date with the earliest first, and break ties by placing the higher consequence rating first.
Stage 3, find the chains. Identify every case where one item cannot begin until another finishes. State each chain as an ordered sequence, and give me the true start-by date for the first item in the chain rather than for the last. Where a chain would already be late given today's date, say so at the top of this stage in plain language.
Stage 4, loss protocol. Separately from the matrix, write a one-page protocol for the trip itself. What to photograph or copy before leaving. Where each copy should live so that no single lost bag, stolen phone or failed account takes out everything. Who my emergency contacts are and what each of them should be holding. And the first five actions to take, in order, if a wallet or a phone disappears while I am abroad. Write it to be read by someone who is panicking, so short sentences and no preamble.
Stage 5, critique your own work. Review stages one to four as a sceptical reviewer would. Name the three items most likely to be wrong, missing or badly estimated for my specific situation, and say what additional information from me would resolve each. Then list anything you have produced that a reader might mistake for a factual ruling about requirements, cover or medical guidance, and rewrite each one as a question with a source.
Stage 6, write the re-run instruction. Give me a short paragraph I can paste back to you next week alongside my updated statuses, so that this becomes a living document. It should tell you what to recalculate, what to leave alone, and what to flag as newly urgent.
How the AI reads this prompt
Practical examples from different industries
A family with an eleven-week runway and a child's first passport. Stage one produces roughly forty rows including several marked as needing applicability confirmation. Stage three is where the trip is saved — it identifies that the child's passport must be issued before any destination-specific documentation can be applied for, and that the passport application itself depends on obtaining a certified birth certificate copy, which has its own processing time. The true start-by date lands eleven days from today, not the six weeks out the family had assumed by looking at the last item in the sequence. Stage five then flags that the birth certificate lead time is the estimate most likely to be wrong and asks which office issued it.
An older traveller coordinating prescriptions across a two-month trip. The complications field lists three daily medications, one requiring refrigeration, and a pre-existing cardiac condition. The matrix separates the pharmacy question — whether a sixty-day supply can be dispensed at once — from the insurer question about declaring the condition, from the destination-specific question about whether a medication may be carried in at all, each with a different source and a different lead time. The chain analysis reveals that the insurance declaration should be resolved before the trip is fully paid for, since the answer may change what cover is available and therefore what risk is worth taking on non-refundable bookings.
A solo traveller working remotely for five weeks across three countries. Stage two produces distinct connectivity and money rows per country, and stage three identifies a chain nobody anticipates — the replacement bank card needed because the current one expires mid-trip must be requested and delivered to a home address before departure, which sets a hard deadline several weeks earlier than the trip itself. Stage four's loss protocol earns its place here more than anywhere, since a solo traveller who loses a phone has no second device in the party and no one holding a backup copy of anything.
Creative use case ideas
Use the same six-stage structure for a house move or an international relocation. The domains change to housing, employment, schooling, healthcare registration and finances, but backwards-dated deadlines and dependency chains behave identically, and relocation is even more chain-heavy than travel.
Run it for a medical procedure with a recovery period. Pre-operative appointments, medication pauses, time off work, help at home and follow-up scheduling form dependency chains where a missed link postpones the whole thing, and the standing rule against factual medical assertions transfers across unchanged.
Adapt it for a wedding, a large family event or a community festival. The consequence rating becomes genuinely useful when a dozen people are each holding a few rows, and the re-run instruction from stage six turns it into a weekly status check rather than a group chat.
Apply it to a certification exam or a professional licence renewal with prerequisites — coursework, supervised hours, application windows and fee deadlines that must be completed in a specific order, where discovering the order late means waiting a full cycle.
Build one for a product launch, a store opening or a first hire. This is the business version, and the only real change is swapping the six domains for legal, financial, operational, staffing, marketing and technical. The stage-three chain analysis is what most project templates leave out.
Adaptability tips
The six domains in stage one are the main lever. Replace them entirely for a non-travel use, or add a seventh for a travel complication the standard set does not reach — pets, vehicles, professional equipment or a stay long enough to change your legal status somewhere all deserve their own branch rather than a row inside someone else's.
Change the sort key in stage two to change the document's purpose. Sorting by deadline gives you a schedule. Sorting by consequence gives you a triage list for the week you ran out of time. Sorting by owner gives you something to distribute. Run stage two more than once with different sorts — it is cheap, since the underlying inventory does not change.
If you want the output narrower rather than deeper, restrict stage one to two or three domains and keep every other stage intact. The chain analysis and the self-critique are the parts that carry the value, and they work fine on a shorter inventory.
For a shared trip, add a field for who owns each row, and add an instruction to stage six asking for the update paragraph to be written in a form that can be sent to another person. The re-run mechanism then works across a group rather than only for you.
Pro tips
Run stage five twice. The first self-critique catches the obvious weaknesses; asking for a second pass on what the first review missed reliably surfaces a different and usually more interesting set. Diminishing returns arrive at the third.
Before you accept the matrix, pick two rows at random and ask what source you would use to verify this and what would make it wrong. If the answers are vague, the row is decorative and the surrounding rows probably are too.
Keep the whole thing in one plain text file rather than a spreadsheet for the first few weeks. You will be editing statuses and pasting the file back into the conversation, and plain text survives that round trip in a way exported spreadsheet formatting does not.
When you re-run it after verifying items against real sources, tell the model explicitly which facts you confirmed and where. It will stop hedging on those specific rows and can then concentrate its uncertainty flags on what genuinely remains open.
Prerequisites
Confirmed departure and return dates, since every deadline in the output is calculated from them and approximate dates produce approximate deadlines that quietly lose their authority. A settled destination, including transit points, and passport details for each traveller including issuing country and age. Ideally your confirmed bookings from Weeks 3 and 4 of this series, because a booked flight is what makes the countdown real. You should also be comfortable working in a longer conversation and pushing back on individual stages — this prompt produces substantial output and is designed to be argued with rather than accepted. Set aside forty minutes for the first run.
Required tools
A general-purpose AI tool with a reasonably long context window and the ability to hold a multi-turn conversation, since the re-run step depends on pasting an updated document back in. ChatGPT, Claude and Gemini all handle this on paid tiers comfortably; free tiers will work for the first run but may struggle when you return with a full updated matrix. A plain text editor or notes app for keeping the working document between runs. No plugins, no browsing and no spreadsheet software required, though you may want a spreadsheet later once the matrix stabilises.
Frequently asked questions
This produces a lot of output. Is the full six-stage version really necessary?
For a simple domestic trip, no — use the beginner variation. The six stages earn their length when there is at least one dependency chain in play, which in practice means any trip involving a document that has to be issued before another document can be applied for, or a medical step that has to happen before an insurance step. If you read your inputs and cannot see a single dependency, you do not need this prompt yet.
How do I keep the matrix updated without re-running everything?
That is what stage six is for. Paste the model's own re-run paragraph back in alongside your edited statuses, and it will recalculate the deadlines that have moved, leave the settled rows alone and flag anything that has become urgent since. The reason the model writes that paragraph rather than you is that it knows which parts of its output are derived and which are inputs, and derived fields are the ones that need recomputing.
What if stage three tells me a chain is already late?
Take it seriously and verify it immediately at the source, because a late chain is the highest-value thing this prompt can find and it is also the one most sensitive to a wrong lead-time estimate. The real processing time from the actual office may be shorter than the model assumed, and expedited options often exist that a general-purpose model will not know the current terms of. Either way you now know which single phone call to make first.
Can I trust the deadline dates?
Trust the arithmetic, not the lead times. The subtraction from your departure date is reliable; the estimate of how long something takes is a planning figure that needs replacing with the real one from the responsible office as you work through the source column. The correct way to use the matrix is to treat every deadline as provisional until the status field says you have confirmed it against the source, which is exactly what the status field is there to track.
Why does it refuse to tell me whether I need a visa even at this level of sophistication?
The sophistication of the prompt has nothing to do with the model's access to current information, and a more elaborate request does not make an unknowable answer knowable — it only makes a wrong answer better dressed. Entry requirements are jurisdiction-specific, change without notice, and depend on your particular nationality and purpose of travel. What the advanced version adds is not an answer but a deadline for getting one, a named source to get it from, and a rating of what happens if you do not.
Recommended follow-up prompts
The Week 7 in-trip prompts from this series, which assume you arrive holding the loss protocol from stage four — that protocol is what the on-the-ground troubleshooting prompts read from when something goes wrong.
A verification-log prompt, run mid-countdown: paste the matrix back with your confirmed answers and ask for a summary of what is settled, what is outstanding and what has changed since the last run, written as a short status paragraph you can share with anyone travelling with you.
The Week 8 reconciliation prompt, which reuses this matrix when closing out the trip — insurance claims, disputed charges and refund requests all reference decisions recorded here, and a matrix with a filled-in status column is the fastest evidence trail you will have.
Tags and categories
Tags:
countdown planning, dependency chains, backwards planning, risk audit, structured output, self-critique, reusable prompts, advanced prompts, emergency protocol, travel logistics
Categories:
Travel & Trip Planning, Systems & Frameworks
Citations
Anthropic prompt engineering documentation — docs.claude.com/en/docs/build-with-claude/prompt-engineering/overview. Covers the structured techniques this variation relies on, including step decomposition, role assignment and explicit output formatting.
World Health Organization, International Travel and Health — who.int. International reference material on health considerations for travellers, cited here as the type of authoritative source the prompt's health rows direct readers toward rather than as a substitute for a clinician.
U.S. Department of State, Bureau of Consular Affairs — travel.state.gov. Carries official guidance for U.S. citizens on lost or stolen passports abroad and on contacting the nearest embassy or consulate, which is the factual backbone of the stage four loss protocol.
Which of the three should you use?
The three prompts differ in what the reader supplies, and that is the cleanest way to choose between them. The beginner Reality Check asks for almost nothing — a destination, some dates, a passport country — and returns the list you did not have. The intermediate Fine Print Translator asks you to supply a document, and everything it produces is grounded in that document rather than in general knowledge, which is why it can say specific things the beginner version cannot. The advanced Countdown Audit Matrix asks for your full profile, your confirmed bookings and forty minutes, and returns a structure rather than an answer. If you have five minutes, run the first. If you have a policy PDF open, run the second. If you have a departure date and a complication, run the third.
They overlap deliberately at one point, and it is the point that matters. All three refuse to tell you what the requirements are. The beginner version routes you to a source type, the intermediate version routes you to a phone call with prepared questions, and the advanced version puts a verification step at the end of every single row and then audits itself for anything that slipped through. That consistency is not caution for its own sake — it is the recognition that a language model's confidence and its accuracy come apart most sharply on exactly these questions, and that the honest version of this help is a well-organised list of what to confirm rather than a set of answers.
The strongest sequence, if you have time for more than one, is first and third with the second slotted in between. Run the Reality Check the week you book, because it costs five minutes and its whole job is finding the long-lead items while they are still cheap. Run the Fine Print Translator when your insurance and card decisions come up, since those are the ones where reading your own document beats any amount of general advice. Then run the Countdown Audit Matrix once your flights are confirmed, feeding it what the first two taught you — and re-run it weekly, which is the part most people skip and the part that turns three good outputs into a plan that stays true until you leave.
TAGS:
