An AI cover letter generator that can only use what you actually did
A cover letter drafted from the same verified record of your experience as your tailored CV — with every claim in the draft extracted and checked against that record before the letter reaches you.
Cover letters fail in a predictable way. Not by lying — by being about nobody. Three paragraphs that would fit any company, any candidate, any Tuesday. That reads as effort without information, which is worse than brief.
Generative tools make this worse in a specific way, and it is worth naming precisely rather than gesturing at “AI slop”. Asked to write a letter from a job description, a model has exactly two sources: the posting, and its own sense of what a strong applicant sounds like. Neither is you. So it writes the letter a plausible candidate would write — and where the posting asks for something it cannot see in its inputs, it produces the sentence that would be there if you had done it.
Drafted from a record, then checked against it
The letter is generated from the same fact graph parsed from your uploaded CV that everything else here runs on. Then — and this is the part that is unusual rather than the drafting — the finished draft goes through the same claim extractor and verifier the CV audit uses.
Reusing that machinery was deliberate. A second, gentler standard for cover letters is exactly how tools relax in practice: the CV gets audited, the letter gets a disclaimer. Here the letter’s prose is wrapped in the same structure the verifier already knows how to read, so it is mined for claims and checked against your history the same way a CV is.
One revision, at most. If a claim comes back contradicted, or unsupported and serious, the letter is redrafted once. The offending sentence is named, and only two moves are permitted: soften it to what the evidence supports, or cut it. Never “find support for it”. Then the revised letter is re-extracted and re-verified from scratch, because the old verdicts describe a document that no longer exists. The result tells you which of the two happened — clean, or revised.
It can say things your CV had to cut
A tailored CV is a compression problem. One page, one role, the strongest subset — and real material gets cut, not because it is weak but because it did not fit this emphasis.
The letter reads from the full fact graph rather than the trimmed CV. So the project that lost its place to something more directly relevant can still appear in the letter, where there is room for a sentence of context that a bullet point flattens. Verified the same way; simply no longer competing for space.
That is the honest argument for a cover letter at all: it is the place for the thing your CV cannot hold.
What you get
The letter as editable text — edit it, because you know things about your own work that no parsed record contains. A PDF export rendered through the same pipeline as the CV export. And the gate’s verdict, so you know whether what you are reading is the first draft or the corrected one.
Regenerating doubles as retrying. There is no separate retry button because there does not need to be one.
What it will not do
It will not write you a personality. The letter can only recombine what your history contains. If the thing you want to convey is enthusiasm for a company’s mission, that sentence is yours to write, and it should be.
It cannot audit your source material. If your CV overstates something, the letter will faithfully carry it. The guarantee is that nothing new is invented.
It will not fill a gap for you. Where the posting asks for something your record does not support, the letter does not paper over it — same rule as the CV. If you want to address that gap, address it in your own words. Saying “I have not done X, here is the closest thing I have done” is a sentence a model should not be writing on your behalf, and it lands better from you anyway.
Questions
- Can an AI cover letter be honest?
- It depends entirely on what the generator is allowed to draw from. A letter written from a job description and a vague sense of what a strong candidate sounds like will invent, because that is the only way to fill the gap between the two. A letter written from a structured record of what you actually did can only recombine what is in that record — and the claims in the draft can then be checked against it, which is the part most tools skip.
- Is a cover letter still worth writing?
- Where one is asked for, a bad one costs you and a specific one helps. The failure mode is not omission, it is genericness: three paragraphs that would fit any company, which reads as effort without information. A letter is worth writing when it says something your CV cannot — why this role, why now, the one piece of context a bullet point flattens.
- How is this different from asking ChatGPT for a cover letter?
- A chat assistant has whatever you pasted into the box and its own sense of what sounds impressive. It has no record of your career to check against, so it cannot tell the difference between a claim you supported and one it produced. Here the letter is drafted from a fact graph parsed from your CV, and then a separate pass extracts every claim in the draft and verifies each against that graph before you see it.
- What happens if the letter claims something I can't back up?
- The gate catches it and the letter is revised once, automatically, with the offending sentence named and only two moves allowed: soften it to what the evidence supports, or remove it. Never find support for it. If it still does not hold after that single pass, you get the letter with the problem flagged rather than an endless loop or a silent pass.
- Can the letter mention something that isn't on my tailored CV?
- Yes, and this is one of its more useful properties. Tailoring trims your CV down to what fits one page and one role. The letter draws from the full fact graph, so it can surface a genuinely strong fact the tailored CV had to cut — still verified, just no longer competing for space.