How to choose an AI visibility tool: 10 questions to ask in a demo

1. Which engines, and which cost extra?

List the assistants your buyers use: ChatGPT, Gemini, Perplexity, Claude, Google AI Overviews, Copilot, AI Mode. Then check the plan, not the homepage. Several vendors include a few engines and sell the rest as add-ons, or let you pick a fixed number per plan. An engine that costs $50 to $400 a month extra changes which plan is cheapest.

2. What does a plan actually count?

Vendors count different things: prompts, credits, answers, “checks” or slots. Convert everything to answers per month:

answers per month = prompts × engines × runs per month

Then divide the price by answers to compare value. Also check limits on brands, projects, competitors, countries and seats; some plans charge per seat.

3. How are answers collected?

Ask directly. Some tools send prompts through the engines’ consumer apps, as a user would see them; some use the engines’ APIs, whose answers can differ from what consumers see; some combine both. Follow-up questions:

• Is the session logged out and free of personalisation?
• From which country and language?
• If an engine changes its interface, what happens to the data?

4. How often does each prompt run?

Daily, every few days, weekly or on demand. Frequency matters more than it looks: to separate a real change from noise, you need many runs. Check whether the frequency drops after an initial period.

5. Do the numbers come with a margin of error?

This is the question that separates measurement from monitoring. AI answers vary, so “visibility 12%” means little without a range. Ask:

• What is the margin of error on my visibility number?
• Is it calculated over prompts, or over answers? (Over answers looks precise but understates uncertainty; see how to measure AI visibility.)
• When the dashboard shows a change, how do I know it is not noise?

Most vendors do not show a margin of error today. If one cannot answer, treat small movements on its charts with caution.

6. Your prompts, or a database?

Two different jobs:

• Tracking your own prompts tells you how you do on the questions you chose, every day.
• Prompt databases estimate what people ask across millions of prompts, which helps you choose what to track.

SEO suites are strong at the second; dedicated trackers at the first. Make sure you can track your own wording.

7. Does it show the sources behind each answer?

You want every cited page, for every answer, with whether it names you. Sources are where most fixes come from (see source gaps). Ask whether you can sort sources by how often they are cited in answers you lose.

8. Does it help you fix what it finds?

Options range from a list of recommendations to drafted content, agents that act, integrations with your CMS or task tools, or a team that works the queue with you. Ask what a typical recommendation looks like, and whether it says why, with evidence from the answers.

9. Can it tell you whether a fix worked?

Ask the vendor to show you a past change and how the tool judged it. Good answers cover:

• When the engines first read the changed page.
• Which prompts the change targeted, chosen in advance.
• How the change compares with competitors over the same days.
• Whether the effect is bigger than its margin of error.

A chart with a marker on the publish date is a start, but not a verdict. The full method is in how to tell if a fix worked.

10. Can you get the data out?

Check for CSV export, an API, BI or Looker Studio connectors, and GA4 or analytics integration to connect AI visibility with visits and conversions. Also ask what happens to your history if you downgrade or leave.

A scorecard for demos

For current prices and features of the main tools, see the best AI visibility tools in 2026.