AI visibility tools compared: how to evaluate Profound, BrightEdge, Semrush and the rest.

The short answer.

To compare AI visibility tools, ignore the headline score and ask how it is made: which AI engines are checked, where the prompts come from, how often each prompt is re-run, whether the tool counts mentions or citations, and which country it asks from. Two tools can report very different “visibility” for the same brand because they measure different things.

The tools fall into three types: AI-native platforms built for large teams, AI features added to an SEO suite you may already pay for, and self-serve trackers with published prices. For many B2B companies with a short list of buyer questions, a careful human-reviewed test, repeated monthly, answers the question before any subscription is needed.

The same word means different maths.

Before comparing tools, separate four things they count. A mention is your brand name appearing in the answer text. A citation is the answer linking to one of your pages as a source. A position is where your name appears in a list. Sentiment is whether the answer describes you favourably. A brand can be cited without being named, and named without being cited, and the two tell you different things about what to fix.

The vendors use similar labels for different calculations. Reading their own definitions on 7 November 2026:

• Peec AI’s documentation defines visibility as how often your brand is mentioned across AI responses, “like market share in AI conversations”, and reports position and sentiment separately.
• Otterly describes “Share of AI Voice” as the percentage of citations you own versus competitors: a citation measure, not a mention measure.
• Ahrefs describes Brand Radar’s AI Share of Voice as your share of AI responses in a topic set that mention or cite your brand, and adds “estimated impressions” weighted by the search volume behind each prompt.
• Semrush describes its AI Visibility Score as reflecting how often your brand is mentioned in AI answers, measured against top competitors in your industry.

So two tools can report very different shares of voice for you and both be correct: one counting answers that name you, the other links to you, over different prompts. When a vendor shows you a score, ask for the numerator and the denominator in one sentence. If they can’t give it, don’t put the number in front of your board.

What to do this week

Decide which of the four you care about most. For a B2B company, mentions in answers to shortlist questions (“which firms should I consider for…”) usually matter most, because that is where a buyer forms a list. Citations matter most when you want AI answers to quote your pages on how a problem is solved. Our article on why being cited is not the same as being chosen explains the commercial difference.

Why one run tells you very little.

AI answers are not fixed like a search results page. Ask the same question twice and you can get a different list. In research reported by Search Engine Journal on 30 January 2026, SparkToro and Gumshoe.ai ran 2,961 prompts across ChatGPT, Claude and Google’s AI features, with each of 12 recommendation prompts repeated 60–100 times per platform. They found less than a 1-in-100 chance that ChatGPT or Google’s AI would return the same brand list twice, and under a 1-in-1,000 chance of the same list in the same order. The work was not peer-reviewed; it found how often a brand appears across many runs more consistent than its rank.

We saw the same spread in a much smaller way when we tested our own category on 6 November 2026. We asked each engine, in a fresh unpersonalised session, which agencies help B2B companies improve their visibility in AI search in the United States.

One run per engine proves nothing about rank. It does show why sampling matters more than the interface. A tracker that runs each prompt once a week is reporting a coin toss. A tracker that runs it daily and reports the share of runs that name you is reporting something you can trend.

Perplexity’s answer also warned that many “top agency” lists it drew on were written by agencies. Remember that with any comparison, including this one: we sell AI visibility work, so we keep to what each vendor says about itself and recommend a tool type, not a vendor.

Eight questions to ask any vendor.

Use these on every demo call. The answers matter more than the dashboard.

Question 4 deserves a closer look. Peec AI states that it uses browser automation to interact with AI platforms through their web interfaces and that this reflects a logged-out user’s experience. Most other vendors’ public pages do not say. Ask, because a signed-in buyer with memory switched on may see something else again (chapter 9).

The tools, side by side.

Built only from each vendor’s own pages as we read them on 7 November 2026. Plans change often; check before you buy. Where a page did not say, we say so.

Prices are the lowest published tier as listed on 7 November 2026, before add-ons and tax. A 15-prompt tier and a 350-prompt tier are different products: compare cost per tracked prompt, per engine, per country.

Profound vs BrightEdge.

The two products start from different places.

Profound is built for AI search first. Its site lists citation analytics (“sources cited across answers”), a visibility score by platform, sentiment and a “FactCheck” feature to verify the accuracy of AI claims, agent analytics showing how AI bots view, cite and refer to your pages, product visibility in ChatGPT Shopping, and “Prompt Volumes”, which it describes as analysing billions of real user conversations. Its pricing page shows a free trial and a custom-priced Enterprise plan with daily tracking, an API and CSV and JSON exports.

BrightEdge AI Catalyst describes itself as “BrightEdge’s intelligence layer for AI search”, and its dashboards pin into the wider BrightEdge environment. It tracks mentions and citations by engine against competitors, rates sentiment, builds prompts mapped to buyer-journey stages and flags when a competitor takes over an answer you held. It also tests whether answers change by the asker’s role, and compares your page with the page currently cited to produce content briefs. Pricing is not listed.

If your SEO team already works in BrightEdge, first check whether AI Catalyst covers the engines and prompts you need. If AI search is its own line of work with its own owner, an AI-native platform is the more direct fit. Either way, run the same 30 of your own buyer questions in each trial and compare the raw answers, not the scores.

Semrush and Ahrefs: AI inside an SEO suite.

If you already pay for an SEO suite, its AI features are the cheapest place to start. “Semrush vs Profound” is really a comparison of types.

Semrush’s AI Visibility Toolkit, per its knowledge base, costs $99 a month with no free trial. The base plan includes one project, 25 tracked prompts and 10 CSV exports a day, with 50 extra prompts at $60 a month. It names ChatGPT, Google AI Mode and Gemini, reports share of voice and sentiment, and adds prompt research and a site audit for issues that block AI crawlers. Prompt Tracking covers more than 220 countries and territories.

Ahrefs Brand Radar takes a different approach to prompts. Ahrefs says it runs more than 538 million search-backed prompts, built from real queries in its keyword database, so you can see a whole topic without writing prompts yourself. Its FAQ says that index is re-tested monthly on a 90-day reporting window, while custom prompts can be checked as often as daily.

The trade-off

A prompt database tells you about a market; your own prompts tell you about your buyers, and a specialist B2B firm may find few database prompts match how its buyers ask. Use the database for discovery, and your own 20–50 prompts for the number you report. A suite add-on keeps AI data beside your keyword and backlink data, which helps when the fix turns out to be an SEO fix.

Peec AI, Otterly and Scrunch.

These three publish their prices, which makes them easy to trial.

Peec AI lists Starter, Pro and Advanced plans at $95, $245 and $495 a month on monthly billing, with 50, 150 and 350 prompts, three active models, daily tracking and unlimited users on every plan. API access starts at Advanced; CSV exports are on all plans. Its documentation is unusually clear on method: answers are collected through the web interfaces, as a logged-out user would see them.

Otterly lists a Lite plan at $29 a month for 15 prompts, with daily tracking across ChatGPT, Google AI Overviews, Perplexity and Copilot; Google AI Mode, Gemini and Claude are paid add-ons. API access starts at the Standard plan. It also offers a content audit aimed at finding “why AI skips your content”.

Scrunch lists Starter at $300 a month month-to-month ($250 billed annually) with 350 custom prompts, 1,000 industry prompts and three personas, across ChatGPT, Claude, Gemini, Perplexity, Google AI Mode and AI Overviews, and Meta. Beyond monitoring, it offers an “Agent Experience Platform” that serves machine-readable versions of pages to AI agents, and a Data API on its Enterprise plan.

How to trial three trackers in two weeks

1. Load the same 20 buyer questions into each, in the same country.
2. Add the same three competitors.
3. After 14 days, export the raw answers and read ten of them yourself.
4. Check whether each tool’s “mentioned” matches what you read.
5. Keep the one whose numbers you could defend to your finance director.

What no tool can tell you.

Every tracker asks AI engines questions on your behalf. That has limits no vendor can remove.

It is a sample, not a census

No tool sees what your buyers actually typed. It asks prompts you chose or the vendor modelled. If buyers ask differently, it measures the wrong thing precisely.

Your buyer’s answer may be personal

A tracker usually asks from a clean, logged-out session. Your buyer may be signed in, in another city, mid-way through a long conversation, or using an app that remembers past chats. Their answer can differ from the one the tool recorded. Treat tool data as a reference point for trends.

It can’t see the outcome

A mention is not a visit and a visit is not an enquiry. Clicks from AI answers are rare: a Pew Research Center study published on 22 July 2025, tracking 900 US adults in March 2025, found that users clicked a link inside Google’s AI summary on just 1% of visits to pages that showed one. Influence happens without a click, so add analytics and a “how did you hear about us” question. Our guide on conversion tracking and analytics covers that side.

What Google itself now reports

For Google’s own AI features, there is first-party data. Google announced Search generative AI performance reports in Search Console on 3 June 2026, and noted that they had rolled out to all websites worldwide by 31 August 2026. Google’s help page says the report counts impressions in AI Overviews and AI Mode, by page, country, device and date. It does not report clicks; Google says traffic from these features is included in the main Performance report under the Web search type. It is free and real rather than sampled, but covers only Google.

When a manual test is enough.

For many B2B companies, a subscription is the second step. If fewer than about 30 buyer questions really matter, a person can run them across three or four engines in a morning and note mentions, citations and errors. Done monthly, in fresh sessions, from the right country, that is a usable baseline.

A manual test also catches what most dashboards don’t: whether the answer is right. When we asked ChatGPT about our own company by name on 6 November 2026, it said we were worth shortlisting, but it quoted an outdated figure from one of our case studies and ran into conflicting founding years: 2015 on our own profiles, 2011 on older directories. A score would have counted that as a win; a person found two things to fix.

Move to a tool when one of these is true: you track more than about 50 prompts; you need daily data to see the effect of changes; several engines or countries matter; or someone needs the data in a warehouse or dashboard. Our guide on how to check your AI search visibility sets out the manual method in full.

Which tool type fits you.

Answer four questions. The picker suggests a type of tool, not a vendor, and what to check before buying. Nothing you choose is sent anywhere.

Whatever the picker says, start with your question list.

Questions owners ask.

What is the most reliable way to evaluate AI visibility products?

Trial two or three with the same 20–50 buyer questions, the same competitors and the same country for two weeks. Then export the raw answers and read a sample yourself. Keep the tool whose counts match what you read and whose method the vendor can explain in plain sentences.

Is Profound better than BrightEdge?

Neither is better in general. Profound starts from AI answer engines and lists features such as prompt volumes and agent analytics. BrightEdge AI Catalyst extends an enterprise SEO platform into AI search. Teams already using BrightEdge for SEO should test AI Catalyst first; teams treating AI search as its own work may prefer an AI-native platform.

Do AI visibility tools show what my buyers actually see?

Not exactly. Most ask questions from a clean session in a chosen country, while a buyer may be signed in, elsewhere, or in a long conversation with memory on. Tool data is a consistent reference point for trends. It is not a record of any individual buyer’s answer.

How many prompts should we track?

Track the questions a buyer asks before contacting a firm like yours: for most B2B companies, 20–50, grouped by stage (problem, shortlist, comparison, risk). More prompts add cost faster than insight unless you sell many distinct services or in several countries.

Can Google Search Console replace an AI visibility tool?

Partly. Its generative AI performance report shows real impressions in AI Overviews and AI Mode by page, country and device, for free. It does not cover ChatGPT, Perplexity or other engines, does not show which competitors appeared, and does not report clicks separately.

Sources: Prolinkage AI self-test on ChatGPT, Perplexity and Google AI Overview, 6 November 2026. Vendor pages read 7 November 2026: Profound (home and pricing), BrightEdge AI Catalyst, Semrush Knowledge Base (AI Visibility Toolkit), Ahrefs Brand Radar, Peec AI (pricing and documentation), Otterly (home and pricing), Scrunch (home and pricing). Search Engine Journal, 30 January 2026, on SparkToro and Gumshoe.ai research. Pew Research Center, 22 July 2025. Google Search Central blog, 3 June 2026; Google Search Console Help, generative AI performance report.