AI search visibility tools run a list of questions through ChatGPT, Perplexity, Gemini and Google's AI answers on a schedule and count how often your brand is named or cited, next to your competitors. They measure the symptom: whether you are in the answer. They don't tell you why you are not.
I use DataForSEO's API for my own measurements, and I still start every client check by hand. Not out of nostalgia. A tracker gives you a clean line on a chart. The reasons behind that line sit where a dashboard doesn't look: a firewall rule, a page that renders only in the browser, a wrong price in a product feed.
Below: seven tools in one table with prices I checked on their own sites on November 18, 2026, four things every one of them misses, the manual check I run for €150, and how to choose by business size.
The short version
If you are a small business with one site and a dozen questions that matter, you don't need a subscription yet. You need one honest measurement and the fixes it points to. A tool earns its fee once you have something to watch: several product lines, several markets, or content going out every week.
Prices moved a lot in 2026. Profound retired its self-serve plans in September. Semrush folded AI visibility into its main plans. Otterly switched to credits on November 10. Any comparison older than a quarter is probably out of date, including this one by spring.
Seven AI search visibility tools compared
Entry price is the cheapest plan with AI tracking, billed monthly, as each vendor's site showed it on November 18, 2026. Currencies are as the vendor showed them to me from Portugal.
| Tool | What it measures | Assistants | Entry price | Best for |
|---|---|---|---|---|
| Semrush (AI visibility in the SEO + AI Search plans) | Visibility score, share of voice, mentions, cited sources, sentiment, prompt research | ChatGPT, Perplexity, Gemini, Google AI Overviews | €190/mo: Starter, 50 tracked prompts, 1 domain. Free plan to try the platform | Teams already doing SEO in Semrush |
| Ahrefs Brand Radar | Mentions, citations and share of voice from a database of 578M+ prompts, plus your own tracked prompts | ChatGPT, Perplexity, Gemini, Copilot, Google AI Overviews and AI Mode; Claude in custom prompts | $199/mo per platform, $699/mo for all. Custom prompts from $50/mo | Benchmarking an established brand against competitors |
| Profound | Daily prompt tracking, citations, AI shopping, agent analytics, content agents | Up to 10 answer engines | Enterprise only, custom quote. Self-serve plans retired in September 2026 | Large brands with a team to act on the data |
| Peec AI | Visibility, position, sentiment, sources; checks answers against brand facts you enter | 3 models of your choice, more as add-ons | $95/mo: Starter, 50 prompts, fact-check on 5 of them | Marketing teams that want a simple tracker |
| Otterly.AI | Mentions, citations, share of voice, sentiment, page audits for AI readiness | Any supported engine; credits decide how many prompts | $29/mo: Lite, about 16 prompts daily on 4 engines | Small businesses and solo marketers |
| Analyze (tryanalyze.ai) | Visibility plus AI referral traffic from GA4 and content workflows | ChatGPT, Google AI Mode, Perplexity; more as add-ons | $99/mo: Growth, 25 prompts daily | Teams that want to tie AI mentions to visits |
| DataForSEO AI Optimization API | Raw data: live AI answers, a database of LLM mentions, AI search volume | ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews | Pay as you go: about $0.0006 per live answer plus the model's own cost | People who build their own measurement (what I use) |
Two things the table can't show. First, the prompt count is the real price: 25 questions tracked daily is enough for most small businesses, 15 often isn't. Second, "tracks ChatGPT" can mean the API model or the consumer app with web search, and their answers differ. Ask which one before you pay.
What every tool misses
None of this is a flaw in one product. It's what a dashboard can and can't be.
Small sites fall out of the database
Database tools build their prompts from search volume. Ahrefs says it plainly on its own page: the index works best once your brand already appears in AI search. A B2B supplier whose buyers ask long, specific questions won't find those questions in any database. You have to write them yourself, and that is judgment, not software.
A mention is not a correct fact
Most trackers count whether you are named and whether the tone is positive. Being named with the wrong price, the wrong spec or a discontinued product counts as a win. Peec has started checking answers against facts you enter, which is the right direction, but it can only check the facts you thought to write down. The errors I find come from where nobody looks: product feeds, PDF datasheets, cached old pages, near-duplicate product variants. On catalogs this costs more than invisibility.
Brand questions inflate the score
Ask any assistant about a company by name and it will usually find it. If half your tracked prompts contain your brand, your visibility looks great and means nothing. The useful question is the one a buyer asks before they know you exist. A tool will track whatever you give it.
Zero looks the same for every cause
When you score zero, a tracker can't tell whether the bot was blocked by a firewall, saw an empty JavaScript page, or read your page and preferred a competitor. In 34 free checks I ran in July 2026, 6 of the first 30 sites blocked AI bots outright, mostly through a Cloudflare default their owners didn't know about. Some tools include a site audit: check that it fetches pages as the AI bots do, not as Google.
Where tools beat a person: repetition. Answers change from run to run, and a tracker asks the same question every day without getting bored. That is worth paying for, once you have the right questions.
The manual check I run for €150
No call. The client sends the site address and two or three main services. Two days later they get a PDF. This is the method.
Write 10 buyer questions without the brand name
Built from the services, in the buyer's words: "who makes X for Y", "best Z for a 20-person company". Plus one control question with the brand name, only to confirm the site is indexed at all.
Ask each question twice in three assistants
ChatGPT, Perplexity and Gemini, with web search on. Ten questions, two runs, three assistants: sixty answers. One answer is an anecdote. Sixty is a measurement.
Record who is named and what is said
For every answer: is the client named, who is named instead, and is anything said about the client wrong. The list of competitors named instead is often the most useful page of the report.
When the answer is zero, find out why
Fetch the pages as GPTBot, ClaudeBot and PerplexityBot, several pages, three tries each with pauses. One 403 can be a rate limit. Three in a row is a block. An empty page means the content lives in JavaScript.
Report N of M and five fixes
"Named in 4 of 60 answers" plus five fixes in priority order. Not "ChatGPT doesn't know you": that is a feeling, not a number.
Why the second run matters: a real case
In October 2026 I was checking an optical retailer. A single run through an AI search tool said nobody makes astigmatism glasses in one day. It would have been the headline finding of the report. Before sending it, I asked the same question in ChatGPT with web search, the way the client's buyers ask. The client came first. One run nearly put a false finding into a paid report. Since then nothing goes into a report until it holds across runs and in the assistant the buyers actually use.
Tool or manual check: how to choose by business size
Prices from the table above. The order matters more than the brand: questions first, then measurement, then a subscription.
Small business, one site, under 15 questions that matter
A manual check every quarter. If you want something weekly, Otterly Lite at $29/mo is the cheapest option with daily tracking.
Growing company, several services or markets
Peec Starter ($95/mo), Analyze Growth ($99/mo) or Otterly Standard ($189/mo), after one manual check to pick the right questions.
Already paying for Semrush or Ahrefs
Try their AI module before buying another tool. Ahrefs fits benchmarking against competitors; Semrush fits teams that live in its SEO reports.
Large brand, many markets, a team to act on the data
Profound, on a custom quote.
You build your own reporting or run an agency
DataForSEO's API: you pay per answer and own the method.
Whatever you pick, the questions matter more than the tool. A good tracker on the wrong questions is an expensive way to feel fine.
The tools got good in 2026. They count faster and cheaper than I can. What they can't do is decide what to count, or explain why the number is zero.
Measure by hand first. Fix what the measurement shows. Then buy a tool to watch it.
Want the manual check on your site?
€150, no call: ten of your buyers' questions, twice each in ChatGPT, Perplexity and Gemini. In two days you get a PDF: who is named instead of you, which facts about you are wrong, and five fixes in priority order. The fee is credited to the audit.
Sources
- Semrush, Plans & Pricing, checked November 18, 2026
- Ahrefs, Brand Radar, checked November 18, 2026
- Profound, Pricing (self-serve plans retired September 2026), checked November 18, 2026
- Peec AI, Pricing, checked November 18, 2026
- OtterlyAI Help Center, plans and pricing, updated November 10, 2026
- Analyze, Pricing, checked November 18, 2026
- DataForSEO, AI Optimization API, checked November 18, 2026
- Author's checks: 34 sites (July 2026), optical retailer (October 2026)
