Give Prompt Fit a prompt and a URL. It reads the page the way a model does, decides whether a real answer is on there, and returns Strong, Partial or Weak with the specific edit that would move it up a grade. It is a read on the page, not a ranking check — we do not query ChatGPT or Perplexity.
“How much does Acme cost for a team of ten?”
The page lists three plans with monthly pricing, but never says whether seats are priced individually. A partial answer is possible; a quotable one is not.
A page can be the top result for “acme pricing” and still leave an assistant unable to state a price. Keywords got you found. Answers get you quoted.
“Flexible pricing built to scale with your team.” Ranks. Reads well. Contains no number, no unit, and nothing an assistant can repeat to the person who asked.
“Pro is $49 per seat per month, billed annually. A ten-person team pays $490 per month.” Still reads well. Now it can be quoted verbatim, with your name attached.
The grade on its own is just a score. What makes it useful is that it comes with the edit that moves it up.
Nothing on the page addresses the question. Answering it means writing something new.
The page circles the question. Usually one missing fact or one sentence in the wrong place.
A real answer is on the page, stated plainly enough to be quoted rather than paraphrased.
Pricing scales with your team. Choose the plan that fits and add seats as you grow — see our plans for details.
Partial — the basis for the price is never stated
Pricing scales with your team. Pro is $49 per seat per month, billed annually — a ten-person team pays $490 per month. Add seats at any time.
Strong — answerable and quotable
The editOne sentence, in the paragraph that already existed. Most promotions from Partial look like this — the page was close, and something specific was missing.
Runs are counted monthly and reset with your billing period.
| Plus | Pro | Agency | |
|---|---|---|---|
| Prompt Fit runsOne prompt tested against one page. | 90 / month | 270 / month | 900 / month |
| Saved to the reportSo the verdict is there when you reopen it. | ✓ | ✓ | ✓ |
| Every prompt you have testedA report accumulates them rather than overwriting. | ✓ | ✓ | ✓ |
Type the prompt a customer would actually type. You get Strong, Partial or Weak, the reason, and the edit that would move it up.
No. Prompt Fit reads your page and judges it against the prompt directly. Nothing is queried, so the verdict is about what you published rather than what a model happens to remember — and it does not move because someone retrained something.
No. It cannot tell you whether you will be cited, because that depends on every other page in the world. It tells you whether a quotable answer is on your page, which is the part you control.
Whether an assistant could quote you. Partial usually means the page implies the answer, or states it a long way from where the question is asked. Strong means it is there, plainly, in a form that can be repeated rather than paraphrased.
Yes, and this catches people out. AI answers come in sizes, and each has a citation budget — roughly 150 tokens for a short snippet, 350 for a summary, 600 for an in-depth reply. We check your answer against all three. An answer that only fits the longest tier is one that never appears in the short replies, which are the most common kind. For a quick length-only check without running a full report, try the free Token Budget Calculator.
That is the single most common Partial. Nothing is missing; the page is asking a system to assemble the answer from three separate spots, and it will not. Consolidating those into one passage under a heading that matches the question is usually the whole fix. You can check whether any of your sections already run long enough to risk this with the free Content Chunking Checker.
The question a customer would type, in their words rather than yours. If you do not have a list, Prompt Discovery builds one from the page — the close-to-winning tier is where testing pays off fastest.
Yes. Optimizations drafts the consolidated answer from content already on your page. Anything the page never stated — a price, a date — comes back as a marked placeholder for you to fill rather than as a guess.
One prompt against one page. Three prompts on the same page is three runs. Re-running a prompt you have already tested on an unchanged page returns the saved result and does not spend another.
Yes. Results accumulate rather than replacing each other, so a page carries the whole set of questions it has been held to.
Prompt Fit tests a prompt you already have. Prompt Discovery finds the ones you have not thought of. Most people build the list with Discovery and test the commercially important ones with Fit.