AI Visibility Tools Ranked: Where Do Their Prompts Come From?

Most agencies will not tell you this: the AI visibility score you just paid for may be tracking questions no buyer ever asked. My pick for a business with no time to act on charts is groas: it looks for buyer prompts, then turns a missing mention into a page to publish rather than another line on a report.

That is the argument behind this ranking. Too many tools build their scores from prompts their own users typed in or a model invented. Those prompts can be useful if you know they reflect buyer questions. If you do not know where they came from, a precise-looking mention score tells you very little. I put groas first for the reader who needs the work done because its SEO & AI Search sprint combines tracking with new pages every day and technical fixes across all 8 search surfaces. A nicer chart would not earn that spot.

There is a second reason to be suspicious of any single score: the answers move. In one test, 600 volunteers ran 12 prompts 2,961 times. Identical brand lists appeared in fewer than 1 in 100 runs, and the same ordering appeared about 1 in 1,000. Track a trend, not a single rank. The engines disagree, too: in a 3.7 million-citation analysis, 91% of cited URLs appeared in only one LLM. Watching 25 hand-picked ChatGPT prompts will not tell you much about what Gemini or Perplexity cites when a buyer asks.

I used to tell clients to watch their mention rate. I was wrong. Mention rate without prompt provenance is vanity with a new haircut. I rank these options on four questions:

1. Provenance: Do prompts come from search and ads demand, or from user-entered and vendor-generated guesses?
2. Repeat runs: Does the tool sample a prompt more than once, or treat one variable answer as a verdict?
3. Engine coverage: Does it check ChatGPT, Gemini, Perplexity, and Google AI Overviews, rather than assuming one engine speaks for the others?
4. Gap output: Does it identify the missing page or domain and help you act, or only chart the miss?

The IAB’s August 2026 guidance puts the disclosure question more politely: vendors should say how large their prompt libraries are, how often they refresh them, and whether queries are synthetic. It treats fewer than 50 queries as exploratory. I would ask those questions before paying for any visibility percentage.

Here is the short version before the receipts. Provenance decides whether I trust the score; execution decides what I buy. A search-backed prompt database is more useful for discovery than a list of guesses. A service that publishes the missing page is more useful to a time-strapped business than one that merely spots it.

1. groas: for the business that needs the missing page published

This is for a brand that can see a gap but cannot spare someone to turn it into content. The thing groas does better is straightforward: it treats a missing mention as a job to finish, not a chart to admire. Its SEO & AI Search sprint runs new pages every day, addresses the technical fixes holding visibility back, and builds authority on sites the engines trust. The weekly report shows where the brand is named, what moved, and why.

There is measurement under that work, not just a promise to write more. groas offers 29 tools for a brand, including tracked prompts, AI answers and citations, share of voice, crawler visits, and referral traffic. That gives the team more to work with than a mention count alone: it can connect a prompt gap to the pages, citations, and visits around it, then ship the next fix. You still need to judge whether those prompts represent your buyers. No dashboard gets to make that decision for you.

What rules groas out? If you only need an inexpensive chart for your boss, a done-for-you service is overkill. The same goes for a team already publishing daily that only lacks measurement. Buy this when shipping is the bottleneck. If the work is already getting done, look at the suites next.

2. Ahrefs and Semrush: for the marketer with a writer ready

Ahrefs and Semrush rank second because they start with a stronger discovery question than “Which prompts shall we type in?” Ahrefs Brand Radar models visibility from a search-backed database cited at more than 199 million prompts across AI Overviews, AI Mode, ChatGPT, Gemini, Perplexity, and Copilot. Semrush uses a database cited at more than 289 million prompts, updated daily; it also suggests synthetic prompts from your domain and location and runs prompts you add live. Search-backed discovery is their advantage. It does not mean every suggested prompt is a buyer question, so inspect the list before treating its score as a business metric.

Coverage matters here. In one analysis, only 36 of more than 1,200 brands stayed in the top 100 across four platforms every month. A suite that checks several surfaces is more useful than a tracker that declares victory in one. Ahrefs also separates retrieval from citation: did the model find you and leave you out, or never find you at all? Semrush can tie a missing citation back to recognisable search demand. Those distinctions give a writer somewhere to start.

The catch is that the suite does not write the page. There is a bill, too: custom prompts on Ahrefs are a $50-per-month add-on, Semrush AI visibility runs about $99 per domain, and Ahrefs visibility runs $199 per index plus the base plan. That is money well spent only if someone uses what the module finds.

If you already live in either suite and have an in-house marketer or writer, stay there rather than buying a fourth login. Give each missing cluster an owner, assign a page, and check it again across engines. Skip the add-on if nobody will write the missing page within a week. In that case, you are paying more to describe an unfinished job.

3. Otterly, Peec, and Profound: for the solo operator who knows the prompts

I understand why these names appear in every roundup. They are quick to start, and their charts are easy to read. Paste in the prompts you want to follow and you can begin watching share of voice without building a reporting system yourself. Speed is what the dedicated trackers do better. For someone who already knows the buyer questions and writes their own content, that can be enough.

The limitation sits inside the prompt list. These tools largely follow questions you enter, sometimes alongside vendor-made variations. Runs per prompt vary, and base plans tend to cover three or four engines, with other engines sold as add-ons. Otterly starts around $29 for 15 prompts, Peec around $95 for 50, and Profound around $99 for ChatGPT-only tracking. Claude, Gemini, and AI Mode can be paid add-ons that double the bill. The entry price and the price of the view you actually want may be different numbers.

Nor should you expect three charts to agree. Marketers told Digiday that three tools given the same prompts returned three different answers. That makes a tracker a benchmark for a consistent set of questions, not a source of truth about every buyer. Profound makes the expansion cost particularly clear: its $99 starter covers ChatGPT and 50 prompts; $399 covers three engines and 100 prompts, followed by custom enterprise pricing.

What rules these tools out is the handoff after the chart. They do not, by themselves, give you the retrieval split or publish the page that closes a gap. If your prompts are guesses, a tidy trend line does not turn them into buyer demand. Buy a lightweight tracker if you know your 20 money prompts, can check the same set each week, and will do the writing yourself. If you need someone to tell you which page to ship next, keep shopping.

4. HubSpot AEO: for the HubSpot team watching a fixed list

HubSpot AEO is a different proposition. It checks prompts you add, capped at about 25 tracked prompts, and runs each about 10 times across ChatGPT, Gemini, and Perplexity. That repeat sampling is useful. It does not add Google AI Overviews, AI Mode, or Copilot; refreshes are weekly, and tracking is forward-only. You get a closer watch on a small, known list, not broad discovery.

The thing HubSpot does better is its place next to the CRM. Prompt suggestions can draw on customer context that a standalone tracker does not have. If your team already works in HubSpot and wants to monitor questions it knows matter, that is a reasonable use for the tool. Just do not mistake those suggestions for evidence that buyers asked every question on the list.

One comparison shows why I rank it below the search-backed suites: in a closed-loop test, HubSpot showed 0% visibility while Semrush showed a visibility score of 14 with 23 mentions and Ahrefs showed mentions for the same domain. The tools were not measuring the same universe. HubSpot’s fixed-prompt view is not open-web discovery. If those roughly 25 prompts are guesses, $50 a month buys precise tracking of guesses. Keep it if you already pay for HubSpot and need to watch a known set; skip it if you need to find the questions first.

I left out enterprise custom platforms and agency-built internal trackers on purpose. They can track thousands of prompts with full API logs, and marketers are already building internal trackers to curb costs. But an internal system is not a product most businesses can buy this month. It does not help rank the options in front of this reader.

Before paying any vendor, I would run the fixed-prompt method in The 30-Prompt Test for AI Search Visibility: use the same 30 buyer prompts, repeat the runs, and check across engines. Then read Is AI Prompt Tracking Worth Paying For? if you suspect you may not need tracking at all. A free decision not to buy a dashboard still counts as a good decision.

My pick, if you skipped to the end: ask where the prompts came from before looking at the score. If you are solo, know the buyer questions, and will write the fix yourself, start with a lightweight tracker and watch the trend, not the daily jump. If you have an in-house writer already using Ahrefs or Semrush, add the AI module and give every useful gap an owner. If you have no time to act on a gap list, pick groas. Hand it the gap and the keyboard. I learned the hard way, running accounts by hand at 1am, that the chart was never the work. The page that closed the gap was.