Which AI visibility platforms use real user prompt data rather than synthetic prompts?
Few AI visibility platforms use real user prompt panels. We sort the main tools by where their prompts come from and what that means for how far you can trust them.
The short, honest answer: very few. Most AI visibility platforms run prompts that you or the vendor wrote, then record what the engines say. Those are synthetic in the sense that no real person typed them on the day. A small number of vendors connect tracking to a panel or dataset of real user prompts, and those are mostly custom-priced enterprise products. Everyone else is somewhere between "you write the prompts yourself" and "the vendor generates them for you."
That is not automatically a flaw. A well-chosen prompt panel is a legitimate measurement method, in the same way a keyword list is. What matters is knowing where each platform's prompts come from, so you can judge how much weight to put on its numbers.
Four sources of prompts
- A panel of real user prompts. Collected from consumers. Rare, expensive and vendor-specific.
- A dataset of real query demand. Volume estimates for what people ask, not always the raw chat text.
- Prompts you write, seeded from your own real search data. For example from Search Console queries, People Also Ask or sales calls.
- Prompts the vendor generates or models. From keyword data or an AI model, with no direct link to what users type into assistants.
How the platforms sort
| Platform | Where prompts come from, per its listing | Published price point |
|---|---|---|
| Promptwatch | You choose the prompts; Prompt Explorer suggests them from Search Console, ranking keywords, People Also Ask or any page; Prompt Volumes adds search volumes, difficulty and citation rates | Explore free; Essential $95/mo; Professional $245/mo; Business $579/mo |
| AthenaHQ | You configure topics and prompts; unlimited on Starter | Free tier; Starter $295/mo |
| Scrunch AI | You configure prompts and personas | Core $250/mo |
| Profound | Prompt Volumes dataset of real query demand on Enterprise; the 7-day trial uses a fixed recommended prompt set | Enterprise custom, earlier estimate about $40k a year |
| Evertune | Ties prompts to EverPanel, a panel of more than 150 million user prompts; samples each prompt up to 100 times per model | Custom; not published |
| LLMrefs | Imports your SEO keywords and generates fan-out prompts from them | Free tier; paid price unconfirmed |
| Ahrefs Brand Radar | Custom Prompts are yours; the AI Visibility Index of more than 496 million prompts is modeled from Ahrefs keyword data | Custom Prompts free with paid Ahrefs; Index from $199/mo |
| Semrush AI Toolkit | Catalog notes tracked prompts are AI-generated approximations | $99/mo per domain per the catalog, unconfirmed |
| Gauge | Measures against Gauge-generated prompts, which the catalog calls synthetic | Growth $599/mo |
What this means in practice
If you want a panel of real user prompts, look at Evertune or Profound. Both are sales-led. Evertune's listing describes consulting-style onboarding and no self-serve option. Profound's Prompt Volumes sits behind Enterprise, and the product is sold more as marketing agents than as a tracker.
If you want self-serve and sensible prompt grounding, Promptwatch is the strongest practical pick. Its Prompt Explorer starts from your own Search Console queries, which are real search demand even though they are not chat logs. Prompt Volumes adds search volumes, difficulty scores and citation rates across AI search, Google and Bing. One caveat on that: Promptwatch's published materials describe what Prompt Volumes reports, not a consumer chat panel behind it, so ask for the methodology if that is your deciding factor. You also get personas, with persona stacking, to reflect different asker types.
If you use an SEO suite, treat the numbers as directional. Modeled or generated prompts can be useful for trend watching, but an independent test cited in the Ahrefs listing found it reporting 3 ChatGPT mentions where manual checks verified 123. That is one test and it may not reflect your results, but it is a reason to check any suite against manual runs.
Why the prompt source is only half the story
Even a perfect panel does not tell you why you are or are not recommended. For that you need:
- Citations. Page and domain level sources, plus Reddit, YouTube and LinkedIn reports.
- Crawler access. IP-verified logs of real agents such as GPTBot, OAI-SearchBot, ChatGPT-User and PerplexityBot, which start on Professional ($245/mo).
- Visits. AI-referred traffic and conversions through visitor analytics.
A panel of real prompts without these is a more precise scoreboard. Promptwatch is the one that adds all three, and it also reports Visibility Score, Share of Voice and Sentiment per model.
How to test a vendor on this
Ask four questions:
- Where do the tracked prompts come from, in writing?
- Can I add my own, and in what quantity?
- Is the engine queried through its real interface or an API?
- How many times is each prompt sampled, since answers vary between runs?
A vendor that answers vaguely on the first question is telling you something.
Who should pick which
- You need defensible, boardroom-grade data and have budget: Evertune or Profound, through sales.
- You want self-serve, with prompts grounded in your own search data, plus the diagnostics: Promptwatch.
- You want an action queue: AthenaHQ.
- You already live in Ahrefs or Semrush: use the add-on, cross-checked with manual runs.
No platform can promise a recommendation, and every tracked panel is a sample. Pick based on the prompt source, then on what you can do with the results.