An evidence-based AI resume builder generates resume language only from career history you have supplied and can stand behind. A general text generator produces plausible resume language from a prompt, whether or not you did the thing it describes.
The two ways an AI can write a resume
Every AI resume tool starts from a large language model, and a language model is very good at producing text that sounds like a strong resume. That is the problem. Asked to “improve” a bullet about running a project, a model will happily add a budget figure, a team size and a percentage improvement, because resumes usually contain those. Nothing in the model checks whether they are true.
An evidence-based builder puts a constraint in front of the model: the source of truth is a body of material the candidate owns — old resumes, project notes, performance reviews, certificates, a LinkedIn export — and the output has to be supportable by something in that material. If the evidence says you led a migration but not how large it was, the resume says you led a migration. It does not guess a size.
What “evidence” means in practice
Evidence is anything that records what you actually did. In FilterProof the sources people commonly add are:
- Every previous version of a resume, including ones aimed at very different roles.
- Project write-ups, design notes, incident reports, internal wiki pages you wrote.
- Certificates, transcripts, licences, training records.
- Performance reviews and written feedback.
- LinkedIn profile exports.
- Plain notes: “in 2019 I owned the on-call rotation for the payments service.”
These land on a private Memory Board as individual pieces of evidence. When you target a job, the builder works from the board, not from a blank prompt.
Why the distinction matters
Interviews test the resume
A fabricated bullet is a question you cannot answer. Interviewers ask about the most specific claims on the page, so an invented metric or an inflated scope tends to surface in the first conversation. Writing only from evidence means everything on the resume is something you can talk about for ten minutes.
Background checks and references
Titles, dates and employers are routinely verified. Claims that drift from the record are a risk that outlasts the application.
Your real experience is usually enough
The more common problem is not that candidates lack experience; it is that the relevant experience is scattered across old documents and was never written up in the language the target job uses. An evidence-based builder is built to recover that material and rephrase it accurately, which is a different job from inventing new material.
What an evidence-based builder does not do
- It does not add skills, tools, or credentials you have not shown evidence for.
- It does not attach numbers to accomplishments unless the numbers are in your evidence.
- It does not promise that an applicant tracking system will rank you, or that a recruiter will call. Nothing can.
It will, however, tell you where the gaps are between your evidence and a job description. That is the honest output: here is what the posting asks for, here is what you have shown, here is what is missing. The candidate decides what to do about a gap — add evidence they forgot, or accept it.
How FilterProof implements it
- Collect. You add evidence to your Memory Board. Anything you have actually done is fair material; nothing is required to be a polished resume.
- Match. You paste a job description, or find one with Find Fit. FilterProof compares the posting’s requirements to your evidence and shows the keyword and requirement gaps.
- Generate. FilterProof writes a tailored, ATS-readable resume and cover letter from the evidence that matches. Generated wording is checked against your evidence so it does not claim experience you have not shown.
- Review. You read it, edit it, and send it. Nothing is submitted on your behalf.
If you want to see how the matching step works on its own, the ATS resume checker page describes it, and how to tailor a resume without inventing experience walks through the manual version of the same discipline.
Frequently asked questions
Does an evidence-based resume builder still use AI?
Yes. The language model still writes the sentences. The difference is what it is allowed to write from: only material the candidate has supplied, with generated wording checked back against that material rather than accepted because it reads well.
What if my evidence is thin for a particular job?
The honest answer is that the resume will be thin for that job too, and FilterProof will show you the gap rather than paper over it. Often the fix is evidence you forgot to add — an older resume, a project you did not think counted. Sometimes the fix is that this is not the right posting.
Is FilterProof free to try?
Yes. The free tier builds one tailored resume and cover letter from your evidence. You create an account first, because the builder needs your evidence saved somewhere before it can compare it to a posting. Details are on the pricing page.