Free prompt difficulty checker
How hard is it to show up for this AI prompt?
Enter your domain and the prompt you want to be visible for. We sample the AI engine’s answers from your country, read every page it used, and score how closed the answer is and how far you are from getting in.
Free, no card, no account. Each check runs on one engine:
- ChatGPT
- Perplexity
- Gemini
- Google AI Mode
- Google AI Overviews
What you get
A score you can act on.
We turn every gap into named work: the pages and searches that would close it, in the order that pays off fastest.
Two numbers, every part shown
Prompt Lock says how closed the answer is to anyone new. Your Difficulty says how far your domain is from the entry bar. Both run from 0 to 100.
From sources to the answer
The pages the engine consulted, the ones it cited and the brands it named, and the stage where your site drops out.
Named and cited, scored apart
Being recommended by name and being linked as a source move independently. Pick the one you want, or both.
Score bands
Prompt Lock and Your Difficulty both run 0 to 100. The cut points match the keyword difficulty bands SEO tools already use.
| Score | Band |
|---|---|
| 0–14 | Very easy |
| 15–29 | Easy |
| 30–49 | Moderate |
| 50–69 | Hard |
| 70–84 | Very hard |
| 85–100 | Locked |
How the score is built
From a sample of answers to the work that closes the gap.
Like keyword difficulty paired with a personal version, the tool returns Prompt Lock (0 to 100, how closed the answer is) and Your Difficulty (0 to 100, how far your domain is from it). Being named in the answer and being cited as a source are scored separately, because different things drive them.
Sample real answers, not one
Two answers to the same prompt rarely name the same brands, so one screenshot proves nothing. We sample the engine you pick, from your country, across same-intent rewordings of the prompt, and measure how often each brand is named and each site cited, with an interval around every share.
Prompt Lock: how closed the answer is
How many slots the core brands hold, how well known those brands are, how much the engine decides from memory before reading anything, and how much of what it reads sits on vendors’ own sites. It is the same for every domain.
Your Difficulty: how far you are from the entry bar
The entry bar is the weakest brand that still makes the target share of answers. We compare your support on the pages the engine reads, your place in the searches it runs and your rankings around the topic with that bar.
What it takes
Every gap maps to named work: the robots.txt line to change, the most-read pages that leave you out, the searches behind the prompt where you don’t rank, the page type to publish. A band is shown only when it holds up under resampling.
The weights behind both scores are first priors from published studies, not yet fitted on outcomes, so read the parts as much as the total.
Questions
Questions worth asking
What the score measures, and what it will not claim.
Why do I get a different answer when I ask ChatGPT myself?
Answers vary from one ask to the next, and a logged-in account adds personal context. The score describes a logged-out user in your country, measured over a sample of answers, so it estimates how often you appear rather than what one answer said.
Why is being named scored separately from being cited?
They move independently. A publisher wants its page linked as a source; a product brand wants to be recommended by name. Engines also differ: ChatGPT cites domains far more often than it names them, Gemini the reverse. The tool picks your main goal from what your site is, or you can choose named, cited, or both.
What is the funnel from sources to the answer?
Before it answers, an engine consults a set of pages, cites some of them and names some brands. The report shows each stage and where your site drops out: never consulted, consulted but not cited, or cited but not named. The plan is written for that stage.
What does "Blocked" mean?
Your robots.txt stops the crawler that engine uses for its answers (OAI-SearchBot for ChatGPT, PerplexityBot for Perplexity, Googlebot for Google’s AI features), so it cannot cite your pages. It is usually the fastest fix, but whether to lift it is your policy call.
Can I compare a score across engines or countries?
No. Each engine builds answers differently and cited sources barely overlap between countries, so a score only holds for the engine, country and language it was sampled in. Run a check per engine.
How reliable is the number?
Each score comes with a likely range and a check of how often it lands in the same band when the answers are resampled. When the band does not hold, the tool says so instead of pretending precision. The weights are priors from published research and will be refit on tracked outcomes.