Would your cover letter work for any company?
Paste a cover letter. One scan redacts every passage that could only have been written to this employer, then scores what remains. No account needed for the first one.
We don't store your letter — it's analysed and thrown away. One free scan without an account.
Paste the job ad too (optional). With it we measure your letter against real postings for this role instead of a reference phrase list — it makes the Anchor score meaningfully more accurate. We store word counts from the posting, never the text.
The four kinds of sentence
Names a decision, product, or fact that fits only this employer. Survives the swap test.
A number, named result, or dated event an interviewer could probe.
Would survive unchanged at any competitor.
A claim with nothing behind it a reader could check.
Reading the Anchor score
| Anchor | What it usually means |
|---|---|
| 0% | The letter never engages with the employer at all. |
| 1–12% | A name-drop or two; nothing that survives the swap. |
| 12–35% | Healthy — real engagement without displacing your evidence. |
| >35% | More about them than about you — research replacing evidence. |
Two letters, measured
| ChatGPT · generic prompt | ForgeLetter · real profile | Your letter | |
|---|---|---|---|
| Anchor | 0% | 25% | ? |
| Proof | 25% | 100% | ? |
| Verdict | FILLER | TARGETED | Scan it |
Both example letters are printed in full in the demo above the fold on our homepage; their scores are computed by this exact rubric, not quoted.
What this does and doesn't tell you
What's measured. Anchor is the share of your letter, by word count, that makes a claim about this employer that would not hold for a competitor hiring the same role. Proof is the share of claims about you that carry a number, a named result, or a dated event.
What's established. A 2025 Yale study of 5.5 million cover letters on a large freelance platform found that before AI writing tools became available, a one standard deviation increase in how closely a cover letter matched its job posting was associated with a 15.46% higher probability of being selected. After those tools arrived, that relationship weakened by 51% for callbacks and 79% for offers — employers shifted toward signals harder to fake, such as work history. The same study found time spent editing an AI draft was positively associated with receiving an offer, while over 75% of AI-generated letters were submitted within a minute of generation.
What that means for these scores. The study measured keyword overlap with the posting. That is not what we measure. Keyword overlap is now cheap to produce, which is why its signal decayed. We measure two things that remain expensive: knowing something real about the employer, and having a number you can defend. We think those hold their value longer. We have not proven it.
What is not established. We have no evidence that a high Anchor or Proof score increases your chance of an interview. Nobody has run that study, including us. These scores describe properties of your letter. They do not predict outcomes.
What we'll do about it.We're collecting outcome data. When we have enough, we'll publish what we find — including if it shows these scores don't matter.
FAQ
Do you store my cover letter?
No — unless you tick the optional research box, in which case we keep a copy with names, employers and figures stripped out. Otherwise it's analysed in memory and discarded: nothing written to a database, nothing used for training. If you paste a job ad we keep word counts from it, never the text. If you click "share result" we save the scores for 30 days.
How many scans do I get?
One without an account, two more with a free account — three total. This tool is a diagnosis; the fixing is what our product does. Paying customers get unlimited scans to verify their letters.
Why don't you swap in a real competitor's name?
Because we'd have to guess one, and a wrong guess is worse than no guess — especially outside the US, where a model is likely to suggest a company from the wrong market entirely. Removing the employer-specific content makes the same point and requires knowing nothing about the company.
Is a high Anchor score always better?
No. Above about 35% the letter is more about them than about you, which usually means research replacing evidence. The healthy band is 12–35%.
Is this the same check ForgeLetter letters go through?
No. This is a standalone diagnostic. Our generation pipeline uses its own quality gate — twelve agents, a hallucination check against your profile, and a coverage check. The two are deliberately separate.
How reliable is the score?
Run the same letter twice and the verdict agrees 100% of the time. We test this on 200 letters before every rubric change and publish the number.