AI Showdown: Three Application Engines Compared
WEEK 104 :: POST 4 :: THE JUDGE’S CHOICE
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 Application Engine — Applying Smart, Not Wide."
This is Week 5 of an eight-week series on running a job search with AI. The reader now has assets (Week 3) and a ranked company list (Week 4). This week they go live — and the trap waiting for them is spray-and-pray: hundreds of applications, a trickle of responses, and morale falling with every silent rejection.
There is a delicious irony at the center of this week, and the post should lean into it — plainly, and without gloating, because some readers will have already sent the two hundred generic letters and this post is their way out, not their scolding: AI-generated application slop is now something recruiters actively watch for. An AI prompt blog is exactly the place to say this plainly: the losing move is using AI to mass-produce two hundred generic cover letters; the winning move is using AI for tailoring intelligence — reading one posting deeply, matching it against the reader's real record, and producing an application that could not have been written for any other job. Same tool, opposite outcomes, and the difference is the prompting — which is the thesis of this whole series, and this is the week to argue it in full rather than just assert it. This is the series' thesis week. The post should say, in as many words, what the series contract promises readers: use AI like an analyst, not a ghostwriter — because employers restrict AI in their own hiring for real legal and compliance reasons, because recruiters discard what reads machine-made, and because an application the reader cannot stand behind in the interview was never an asset to begin with.
By this point the reader is also juggling artifacts from four earlier weeks — a spec sheet, a master résumé, a dossier — and keeping that context straight across sessions is itself becoming work. The strongest posts can acknowledge that honestly: structured, reusable prompting is what keeps a multi-week project coherent.
The deliverable the reader should walk away holding: a working application pipeline — a tracked set of live, tailored applications aimed at the top of the Week 4 ranking, with outreach running alongside.
THE SERIES CONTRACT — identical every week; it binds every prompt you design. This series' tagline is its editorial contract: "Use AI like an analyst, not a ghostwriter." It is written for a reader in a market unsettled by AI itself — some readers are searching precisely because AI eliminated their last role. Write with that reader at the table: no AI-efficiency cheerleading, no automation jokes, no promises that AI will "do it for you" anywhere a human hiring decision is involved. And hold one line in every prompt: the AI is the reader's private analyst, coach, and sparring partner — it structures, researches, rehearses, and questions. It does not ghostwrite. Anything a hiring human will read or hear — résumé lines, cover letters, outreach messages, interview answers, negotiation emails, resignation letters — must end in the reader's own words and be true. Prompts should drive toward drafts the reader rewrites and owns, and should say so explicitly. Employers increasingly restrict how AI may be used in their own hiring decisions for legal and compliance reasons, and recruiters increasingly recognize — and discard — material that reads machine-written. A prompt that makes a reader look AI-generated hurts them twice. Posts that ignore this contract should expect to lose the week. Two practical notes. First: a standing “About this series” notice covering these same points is added to every published post automatically at publication — acknowledge the frame in your own voice where your week's prompt calls for it, but do not write a formal disclaimer block of your own, and do not open every post with the same acknowledgment paragraph: outside the weeks whose prompts explicitly carry the series frame, this contract lives in your tone and your prompt design. Second, the framing is POSITIVE: used this way — analyst backstage, reader on the page — AI is an advantage no hiring human will ever hold against your reader. Write like that is true, because it is.
The three prompts should help a reader:
- Decode the job description. What the posting actually says versus what it means: the must-haves versus the wish list, the responsibilities hiding behind buzzwords, the signals about level and team — so the reader decides in minutes whether this posting deserves a tailored application at all.
- Build one tailored application package. A résumé variant cut from the Week 3 master against this specific posting, and a cover letter grounded in the reader's actual specifics — concrete, short, and impossible to mistake for a template. The reader supplies the posting and their real materials; nothing gets invented.
- Run the full pipeline system. The operating rhythm: an application tracker with stages and dates, referral-outreach messages that respect the person being asked, informational-interview requests that are not stealth applications, and a follow-up cadence that is persistent without being unhinged.
At the advanced tier, the strongest version of this week is the pipeline as an operating system — decode, tailor, send, track, follow up, learn — where the tracker feeds back into targeting (which kinds of postings respond?) and outreach runs as a parallel lane rather than an afterthought. That loop is worth reaching for.
A constraint for this week. Response rates, referral-conversion multipliers, and "applications per hire" figures circulate everywhere and verify almost nowhere. No prompt may ask the AI to state response or conversion statistics as fact, and no outreach prompt may have the AI fabricate a relationship, a mutual contact, or credentials the reader does not have. Honest outreach written from the reader's real situation is both more effective and the only kind this blog will teach.
Design the prompts so the AI does what it is genuinely good at: deep reading of one document, precise matching against another, drafting from supplied specifics, and imposing system on a chaotic process. The reader supplies the postings, the master assets, and the truth; the AI supplies the decode, the tailoring, and the discipline.
Series dependency chain, for the Metadata block: Week 5 consumes Week 3's master résumé and positioning narrative (variants are cut from them) and Week 4's ranked dossier (effort goes to the top of the list). Week 5 produces the live application pipeline — consumed by Week 6 (interviews come out of the pipeline) and Week 7 (multiple live processes become negotiation leverage).
Because readers may arrive at this post without having read the earlier weeks, the prompts should work for someone with a résumé and a posting in hand, while making clear the engine runs far better fed by a real Week 3 master and a Week 4 ranking.
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 career topic that touches every industry. The template lists tech startup / retail / freelance as suggested industry examples — those are marked MAY, and adapting them is expected here. A reader tailoring one application to a dream posting, another asking a former colleague for a referral without making it weird, and a third rescuing a pipeline that is all silence are the right kinds of contexts. 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 — response-rate and referral-conversion numbers especially. If the sentence needs a statistic to land, rewrite the sentence.)
## 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: 5` 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 response or conversion statistics as fact, and no outreach prompt invites the AI to invent a relationship, mutual contact, or credential the reader does not actually have. Also confirm every prompt that produces a cover letter, outreach message, or any other text a hiring human will read ends with the reader rewriting and owning the final words (the series contract's final-pass rule) — this is the week that produces exactly the material recruiters screen for AI slop.
A note on this week's scoring, and this one is our fault rather than any of the three AIs'. Gemini judged this week and marked ChatGPT down hard on Template Compliance — 3 out of 10 — saying it "completely omitted all ## markers from its headings." That is not true. ChatGPT's post has all 57 of the required headings, correctly structured, and you can see them in the post itself. Here is what actually happened. ChatGPT returns its posts to us as a Word document. In a Word file, a heading is a style rather than a visible marker, so the "##" characters our checklist asks about simply are not there to see — the structure is right, the symbols are invisible. We passed that file to the judge as we received it, and asked the judge a question that a Word document cannot answer. Two of the three judges in this series opened that file correctly and said so; ChatGPT's own week noted that the document "contains all 57 required Heading 2 sections." Gemini did not, and it has happened in all three of the weeks Gemini has judged. So ChatGPT is carrying a seven-point penalty for a fault it did not commit, and the fault is in how we delivered its work. We are publishing the scores exactly as Gemini returned them, unedited, because correcting a judge's numbers after the fact is not something we will do — but you should read this week's table knowing that ChatGPT's real Template Compliance is a 9 or a 10, which would put it at 64 rather than 57. It does not change who won: Claude's 70 out of 70 leads either way, and that is the first perfect score we have published in any series. One more thing worth your attention, on the winner rather than the loser. Nobody independent read the three posts this week. Our rule is that we bring in a second reader whenever the judge picks its own post to win, and Gemini placed itself last, so that rule did not fire. A flawless score therefore rests on one judge's word alone. We think Claude's post is the strongest of the three; we are telling you plainly that we did not verify it.
About this series. AI is reshaping work — for some readers, it's part of why you're searching at all. We don't pretend otherwise. And the hiring world is wary of AI-written material: many employers restrict how AI may be used in their own hiring decisions, and recruiters increasingly recognize — and discard — machine-written applications. So this series teaches a different approach: use AI as your private research analyst, interview coach, and thinking partner, while every word an employer sees or hears from you stays genuinely, verifiably yours. AI behind the scenes. You on the page.
Scored across seven dimensions by this week's rotating judge. The judge scored all three posts, including its own, with authorship visible.
Claude takes Week 5 :: Job Search Series with 70 of 70, ahead of ChatGPT on 57.
Each dimension scored 1-10 by the judge. These are the judge's own scores, not measured data.
Dimension by dimension
1. Prompt Quality & Creativity — Claude wins.
Claude’s prompts are exceptional, introducing highly inventive and robust mechanisms like a 5-section decoder, a strict "strong/partial/none" evidence map, and a continuity block for cross-session memory. ChatGPT is also excellent, utilizing claim ledgers and state reconstruction to build a highly disciplined system. As Gemini, applying strict scrutiny to my own post, I recognize that my prompts are standard and lack the deep, systemic scaffolding seen in the other two. 2. Content Depth & Accuracy — Claude wins. Claude perfectly frames the philosophical shift of using AI as an analyst, teaching the reader how to extract real signals from vague postings and map evidence ruthlessly. ChatGPT provides similarly deep insights into the mechanics of job targeting. Gemini falls short under the series contract rule. The contract explicitly states: "...must end in the reader's own words and be true. Prompts should drive toward drafts the reader rewrites and owns, and should say so explicitly." Gemini’s Prompt 3 generates an outreach draft but fails to explicitly require the reader to rewrite it, causing a deduction here. 3. Template Compliance — Claude wins. Claude perfectly executes all structural requirements and uses NOT APPLICABLE honestly for citations in Variation 1. Both other posts suffered gross structural failures that should have been caught in the pre-flight check. ChatGPT completely omitted all ## markers from its headings, violating the rule: 'Every template heading written as ##, none bolded instead'. Gemini similarly failed by bolding its headings (e.g., ## Lead) and using nested headings in the breakdown, violating the rule: 'Every template heading written as ##, none bolded instead; prompt breakdown is running text split on :, with no ### headings inside it'. 4. Practical Utility — Claude wins. Claude's prompts produce immediately usable, highly specific artifacts. Its "Pipeline Operating System" logically structures tracking and parallel outreach lanes in a way a job seeker could adopt today. ChatGPT is extremely thorough and actionable, providing excellent guardrails. Gemini loses points under the series contract rule for its Prompt 3 outreach draft (as noted in Dimension 2), and its pipeline manager lacks the concrete field definitions and state-tracking mechanisms present in the others. 5. Engagement & Readability — Claude wins. Claude's prose is sharp, engaging, and perfectly matches the requested WSJ/Forbes register. Openers like "Most people read a job posting the way they read a menu at a restaurant they have already decided to eat at" are brilliant and relatable. ChatGPT reads a bit more technically and densely, resembling an instruction manual in parts, though it remains highly professional. Gemini is accessible but relies on slightly generic hooks. 6. Citation Quality — Claude and ChatGPT tie. Claude provides highly relevant, real-world citations regarding the EEOC, NYC Local Law 144, and the Illinois Video Interview Act to explain why employers restrict AI. ChatGPT equally cites real sources like O*NET and a 2025 hiring-manager survey. Gemini simply cites the assignment itself ("Ketelsen.ai Job Search Series Contract") for all three variations. While not fabricated, it is thin and misses the opportunity to ground the post in external reality. 7. Tier Differentiation — Claude wins. Claude distinctly separates the tiers by cognitive task: reading one document (Beginner), matching two documents (Intermediate), and running a multi-document ongoing system (Advanced). ChatGPT follows a similar, highly effective operational progression. Gemini scales logically but feels more like a progression in prompt length rather than a fundamental shift in operational complexity.
Winner
Claude
is the clear winner this week. It fully understood the thesis of the assignment—using AI as an analyst rather than a ghostwriter—and built structural constraints directly into its prompts to enforce that behavior. The writing is exceptional, the formatting is flawless, and the progression of the prompts provides a masterclass in how to manage a complex, multi-week project using an LLM.
The Honest Counter-Case
While Claude takes the win, ChatGPT deserves massive credit for the sheer rigor of its prompt design. The inclusion of a "claim ledger" in its tailoring prompt is a brilliant way to ensure provenance and prevent hallucination, arguably offering slightly better anti-fabrication guardrails than Claude. Gemini (my own post), while the weakest overall, did an excellent job keeping its beginner prompt remarkably lightweight and accessible, whereas both Claude and ChatGPT demand quite a bit of upfront reading and setup even at the beginner tier.
Takeaway for Readers
The difference between a bad prompt and a great prompt is often just boundary-setting. When you ask AI to "help me apply," it will default to writing generic fluff. When you assign it a specific role (e.g., "hiring-side reviewer"), give it strict evaluation criteria, and explicitly forbid it from ghostwriting, it transforms into an analytical engine that can map your real evidence to an employer's actual needs. Stop asking AI to write for you, and start asking it to read for you.
TAGS:
