Keep the Log Now: It Is the Evidence You Cannot Recreate

What One Entry Should Actually Contain

A useful entry is short: the task, what the assistant produced, what you changed and why, and what it would have cost if you had shipped the first draft unchecked. Four or five sentences, written the same day, beat a detailed memory reconstructed months later when someone finally asks. The point is not volume. It is that each entry can answer a follow-up on its own.

A weak entry says 'used ChatGPT to help with a report.' A useful one says: asked it to summarize twelve customer interviews, it merged two contradictory complaints into one, caught that during a read-through, split them back out and flagged the pattern to a manager who had missed it. The second version is not longer for its own sake. Every clause in it is something an interviewer could ask about.

The cost line is the part most people skip, and it is the part that separates a real entry from a routine one. Not every use had a cost worth noting: plenty of AI work is small and low-stakes, and those uses can stay a single line. Note the cost specifically when the assistant was confidently wrong about something that mattered, a number, a fact, a recommendation, because that is the entry a reviewer actually wants to hear about.

Write It Down at the Time, Not Afterward

Write the entry within a day of the work, not at the end of the week and never during a job search six months later. Specifics decay fast: what a chatbot got wrong, the figure you had to recompute, the sentence you deleted because it read confidently and was false, all of it blurs into 'I used AI a lot' within a month.

A one-line prompt in a note-taking app, filled in properly within a day, holds the detail a resume bullet needs later. Set a recurring reminder if that is what it takes; the habit is the whole system, and a system that depends on remembering to remember tends to stop within a few weeks.

Keep the file somewhere only you can see it: a personal notes app, a private document, not a work tool an employer owns. Ten entries, kept this way over a few months, is a small file that takes minutes a week and pays out at the exact moment a specific claim is worth more than a general one.

Why Nobody Sells You This Advice

No course, template, or coach makes money teaching you to keep a private log, because the payoff arrives months later and there is nothing to purchase at the moment the value would be obvious. That gap in the advice market is not evidence the practice does not matter. The employer side of hiring is standardizing on exactly the kind of question only a log answers well.

In a Robert Half survey of more than 2,000 US hiring managers, published in 2026, 67% said that reviewing AI-generated applications has slowed their process, and running more interviews per candidate (38%) was among the responses employers named ¹. Interviewers under that pressure are not asking whether you used AI. They are asking for one specific instance, and a candidate reciting a log entry sounds different from one reaching for something that happened once, vaguely, at some point.

The same survey found 65% of those hiring managers saying a surge in applications, many of them AI-enhanced, has made candidate skills harder to verify ¹. A written log does not fix an application; it fixes the conversation that follows one, which is where a vague claim actually costs a candidate the offer.

Protect What Needs Protecting As You Go

Strip anything confidential the moment you write the entry, not later when you try to reuse it. Name the task and the shape of the problem rather than the client, the figures, or the internal tool. Confidentiality does not have to end the practice: decide which details your agreement covers, and keep the reasoning behind the decision, which is the part the log needs anyway.

Written this way, the same entry works whether the job you eventually describe it in is covered by an agreement or not. You are not storing evidence that could get you in trouble. You are storing the account you would give anyway, in a form you can actually find again.

Ten Entries Are Enough to Start

Ten real entries, kept honestly, source every AI-related claim you might put on a resume and answer nearly any follow-up an interviewer asks. Naming AI skill on an application is common now: in Handshake's analysis of resumes on its platform, Class of 2026 resumes mention AI skills more than nine times as often as Class of 2022 resumes did, and two-thirds of those mentions now come from majors outside computer science ². A vague 'AI-proficient' line does not meet that bar.

In a June 2026 ZipRecruiter survey of more than 1,000 hiring decision-makers, 74% called AI skills a strong advantage or a requirement outright, and half said they expect candidates to already be practical or advanced users ³. A claim sourced from a real entry meets that expectation; a claim with nothing behind it does not, and what verifying that claim actually looks like is worth reading once you have a few entries to draw on. The log is the raw material for that conversation.

The risk sits on being caught without an answer, not on having used AI. In the closest available study of trust and AI disclosure, a professional whose undisclosed AI use was exposed by someone else was trusted less than one who had disclosed it himself; the scenario was a professional and his client, so it is the ordering that transfers, not the size of the gap ⁴. A log kept over a few months is what keeps a candidate out of that gap, and it costs a few minutes a week to build.

See a sample report >