AI visibility tool buying checklist: 38 questions to ask before you sign

How to use this checklist

Bring this list to every demo call. Ask each question, write down the answer, and ask the vendor to show it on screen rather than describe it. A feature that only exists in the sales deck is not a feature.

The questions are grouped into nine sections. Not every section matters to every buyer. An in-house brand team can skip the agency questions; a team that only wants a mention count can skim the diagnosis section. But read the red flags in each one, because they come from patterns we've seen across the tools in our catalog, and they're the things buyers tell us they wish they'd asked.

1. Engine coverage

Which engines are included on the exact plan I'm buying, not the top tier?
Are Google AI Overviews and AI Mode tracked separately?
Is Claude covered? Copilot? Perplexity?
Can I see answers by country, and if I need it, by state or city?

Red flag: an engine count on the homepage that only applies to Enterprise. Writesonic GEO self-serve plans cover three engines, and the rest require Enterprise. Profound Starter covers ChatGPT only. Neither is hidden, but both are easy to miss.

2. Usage limits and billing model

Is usage metered in prompts, responses, or credits?
What exactly does one credit buy: a prompt, a prompt on one engine, or a prompt on one engine on one run?
What happens when I hit the limit: overage charges, a pause, or a downgrade in frequency?
How often are prompts checked on this plan: daily or weekly?
Do unused credits roll over?

Red flag: a meter that stops the product. Qwairy pauses new analyses when credits hit zero. AthenaHQ charges a credit per AI response, and G2 reviewers on our listing describe using a month's allowance in the first week. Do the math: prompts times engines times runs per month, against the plan's allowance.

3. How the data is collected

Is each engine captured from the real consumer interface or sampled through an API?
How many times is each prompt run per check?
Can I see the full answer text behind a data point?
How do you handle personalization and location?

Red flag: no answer to the first question. Vendors that are transparent will tell you engine by engine. Our listings note that Profound's citation counts have been reported to diverge from manual ChatGPT checks, and that Ahrefs Brand Radar models its prompts from Google keyword data. Sampling is always directional, but you should know how directional.

4. Prompt intelligence

Can I see search volume or demand estimates for each prompt?
Is there a difficulty score?
Does the tool show query fan-outs, the sub-searches an engine runs to build its answer?
Can I tag prompts by topic and by persona?

Red flag: a tool that generates your prompts for you and won't say where they came from. Semrush AI Toolkit's tracked prompts are AI-generated approximations, per our listing. That can be fine for direction. It isn't the same as tracking what buyers actually ask.

5. Diagnosis: why did it change?

Can I see which sources an engine cited for each prompt, down to the page?
Are Reddit threads, YouTube videos, and offsite mentions broken out?
Do you show AI crawler logs, meaning which AI bots fetched which pages and with what errors?
Can I trace a page from crawl to citation?
Is there a trend view showing what changed between checks?

Red flag: "we show citations" meaning a domain list only. A domain doesn't tell you which page to fix. And if the vendor has no crawler data, it can't tell you whether a lost citation was a content problem or a technical one.

6. Attribution

Can the tool show AI-referred visits to my site?
Are those visits tied to conversions?
How is tracking installed: script, tag manager, or integration?

Red flag: "coming soon." Our Brandlight review notes its attribution module was still listed as coming soon in early 2026. If leadership will ask what AI search is worth, buy a tool that can answer today.

7. Turning findings into work

Does the tool produce a prioritized list of actions?
Can it draft content against prompts where I'm missing?
Can it publish to my CMS, with a human review step?
Which CMSs are supported?

Red flag: recommendations so generic they could apply to anyone. Our Semrush listing quotes advice like "improve onboarding" and "lower your prices." Ask to see three recommendations for your own domain.

8. Access, reporting, and agency needs

Is there an API? An MCP server? On which plan?
Can I export to Looker Studio or my BI tool?
How many seats and projects are included?
For agencies: are there client workspaces, custom reports, and white-label options, and on which plan?

Red flag: reporting that ends in a CSV. The most common G2 complaint on our Scrunch AI review is that users export to Excel to build client charts.

9. Contract terms

Is monthly billing available, or annual only?
Is there a free tier or a trial on the plan I'd buy?
Does the contract auto-renew, and are there price escalators?
Are there implementation fees?
How long until the data is reliable enough to act on?

Red flag: multi-year auto-renewals with escalators. Our BrightEdge review documents 12 to 36 month auto-renewing terms with built-in escalators. Our Conductor listing cites implementation fees of $20K to $150K. Profound's public tiers bill annually only, and its listing budgets one to three months of setup.

How Promptwatch scores on this checklist

We ran Promptwatch through the same list, using its published fact sheet. It's the tool that answers the most questions with a yes on a self-serve plan, which is why it leads our ranking.

Section	Promptwatch, as published
Engines	ChatGPT, Gemini, Claude, Perplexity, Grok, Llama, DeepSeek, Mistral, Copilot, AI Overviews, AI Mode on every paid plan; country on every paid plan, state and city from Business and on agency plans
Usage	Fixed limits: 6,000 / 18,000 / 42,000 responses on Essential / Professional / Business
Data method	Real UI monitoring, not just APIs
Prompt intelligence	Search volumes, difficulty scores, query fan-outs, topics, tags, personas
Diagnosis	Citation analytics at page, domain, Reddit, YouTube, and offsite level; citation and prompt trends; Agent Analytics crawler logs with crawl-to-citation path from Professional
Attribution	Visitor analytics with conversions via script or GTM template, from Essential
Action	Unified Actions to-do list; Content Agents publish to Webflow, Framer, WordPress with a review inbox; automated Content Agent from Business
Access	MCP and API on every paid plan; Looker Studio connector on agency plans
Contract	Monthly billing, or annual at 12 months for the price of 10; free Explore tier; 7-day trial on Essential and Professional

Two honest caveats. Brand-plan seats are capped (one on Essential, two on Professional, five on Business), so large teams should look at Business or an agency plan. And crawler logs aren't on Essential; you need Professional ($245/mo) for Agent Analytics.

Scoring your shortlist

Give each finalist a point per checked box in the sections that matter to you. Double-weight sections 5 and 6 if you'll need to explain changes or prove value, because those are the hardest gaps to fill later with a second tool. Then compare the totals against the real monthly cost from section 2.

For the full field, see our full ranking, and for a walk-through of the decision itself, our guide on how to choose an AI visibility tool.

FAQ

Which sections matter most?

For most buyers, usage limits (section 2) and diagnosis (section 5). The first decides what you'll actually pay. The second decides whether the tool can explain anything.

Should I trust a vendor's engine count?

Only for the plan you're buying. Ask which engines are on that plan specifically.

Is a free trial enough to evaluate?

Usually, if you load your real prompts and test one lost mention end to end. A free tier with only one engine won't show you multi-engine differences.