Most Popular AI Visibility Products for SEO, by What They Measure

On 7 November 2026 we read the pricing and method pages of 14 vendors. Two things came out that the ranked lists skip. Most vendors say very little about how their numbers are collected, and the headline percentage is far shakier than the chart makes it look. This post covers what each type measures and cannot see, how to read the number, what the plans cost today and which type fits which buyer.

Where this post sits: it is the category overview. For trackers that watch ChatGPT, read AEO tools that track ChatGPT brand mentions; for Google AI Mode, read the review of the best AI Mode SEO tracking tools; for the method of watching named rivals, read how to track competitors in AI search results.

Our disclosure: CiteVantage is an agency, we sell services and none of these products, and we have no affiliate links or referral deals with any vendor here. We have signed up for some of them ourselves, Semrush and Ahrefs among them, and use their free tools from time to time.

What does an AI visibility product actually do?

It asks AI engines a fixed list of questions on a schedule and records whether your brand is named or linked in each answer. Charts, alerts, competitor columns and client-deck exports are presentation on top of that log.

Nearly every product reports the same five things:

• Mention rate: the share of your tracked prompts where the engine names you.
• Share of voice against a list of competitors you define.
• Citation sources: the URLs the engine used to build the answer, which is the most useful number in the set and the most ignored.
• Sentiment of each mention.
• AI referral traffic, taken from your own analytics because the engines do not report it.

Engine coverage is where plans differ most. Peec AI lets you choose three of six engines on its entry plans, Otterly.ai includes four and sells Claude, Google AI Mode and Gemini as add-ons, and Profound’s free trial covers three engines while its Enterprise tier lists up to nine. Check which engines your plan includes before you compare anything else.

What are the most popular AI visibility products for SEO, by type?

Five types, and the right one depends on whether you are checking, monitoring or reporting to someone else. A brand asking whether it exists in AI answers needs a different product from an agency proving retainer value across a dozen clients.

Free graders run a handful of prompts once and hand back a score. HubSpot’s grader says its result is based on the engines’ training data, which suggests it is not testing what a live search would return, so treat it as a first look and nothing more.

Dashboards are the core of the category. You write the prompts, the product runs them daily, and you get mention rate, competitor share and cited URLs over time. Our comparison of the ChatGPT-focused trackers is in AEO tools that track ChatGPT brand mentions.

Suite modules are the cheapest route if you already pay for the suite. Ahrefs includes prompt tracking in its paid plans, while Semrush sells its toolkit as a separate $99 per domain line, and Google AI Mode coverage varies a lot between them, which we broke down in the review of AI Mode SEO tracking tools.

Profound is the only vendor we read whose pricing page offers nothing between a free 7-day trial (50 prompts, three engines) and a custom Enterprise quote. A third tier called Agency Growth appears only in its FAQ, with 400 credits a month per client workspace and no price.

Frase’s own pricing page says its AI Visibility feature does not auto-fix anything, which is an honest line and the right way round. The fixing layer in this type is a second product on the same invoice.

What does each product measure, and what can it not see?

Each one measures a scripted, anonymous session on a schedule, and whether that session resembles your buyer depends on five settings that most vendor pages state only in part. We went through each vendor’s own pages on 7 November 2026 and wrote down what they say.

Read that table as a list of questions to put to a vendor, because “not stated” is not the same as “done badly”. It means you cannot tell from the page. The five settings, with what we know from our own runs:

• App or API. The two can return different markets for the same engine. In our 500-query Houston study, Gemini through its API named 7.8 firms per answer and the Gemini web app 3.9, across 473 queries both answered in September 2026, and on 56.5% of the 434 queries where both named a firm they named nobody in common. Peec rejects APIs for this reason, and we agree with it.
• Location. Six of the seven products describe country as the finest control. HubSpot’s one-time grader takes a location you type, and a city is among the examples its form gives, without saying how it is used. For a local business the city has to live in the prompt text. In the same study, questions without a state marker such as TX or Texas were answered about the wrong place 4.5% of the time (50 of 1,120), against 1.1% (10 of 880) with one.
• Logged in or out, and memory. A signed-in session with memory answers a different question than a stranger asks, which is why our methodology uses fresh chats with no personalisation carried over. Even that can slip: in the late-October repeat runs behind our November Houston report, ChatGPT’s temporary chat began reading memory before the repeats finished, and we kept only the questions with three clean answers. A tracker running a thousand prompts a day can hit the same problem, and you would not see it.
• One run or many. We cover this properly in the next section, because it is the setting that moves the headline most.
• Named or cited. Being mentioned in the prose and being linked as a source are separate outcomes, and we log both. Ask a vendor which one its percentage counts.

One gap we cannot measure: trackers send a single cold question, while a buyer often arrives three messages into a conversation. We have no figure for how much that changes the answer.

How do you read a tracker’s headline percentage without being fooled?

Treat it as a share of a sample, then ask five questions about the sample before you act on it. If you appear in 18 of 60 tracked prompts the dashboard says 30%, and one prompt flipping moves the number by 1.7 points.

Prompt share

30%

is what a dashboard reports when you appear in 18 of 60 tracked prompts, and that is not 30% of anything a buyer sees.

Photo: This_is_Engineering, Pixabay

The arithmetic is ours and it is the generous version. If every one of the 60 prompts were an independent draw, the 95% interval around 30% runs from about 20% to 42%. Prompts are not independent, since they share a buyer, a market and often a phrasing, so the honest interval is wider. A week-on-week move from 30% to 33% is two prompts changing their answer, and nothing in it should reach a client report.

The five questions, each with what our own published runs showed:

1. Which surface produced it? The Gemini figures above are the example. A “Gemini visibility” number that does not say whether it came from the app or the API describes neither.
2. How many runs sit behind each prompt? In our November Houston report we put a random 100 October questions back to ChatGPT and Gemini twice, in real browsers, with the state named in every question. For the average ChatGPT question, between 26% and 38% of the firms that either of two runs named were named by both. For Gemini the range was 10% to 15%. On 39% to 53% of Gemini questions, two runs named no firm in common. That sample is 63 ChatGPT and 61 Gemini questions in one city, so it shows scale and not a universal rate.
3. From where? If the setting is country-level and your business is a city, the prompt text carries the geography. Check that it does.
4. What counts as a miss? An empty or failed response logged as “not mentioned” inflates the absent count. We hit this in our 181-brand study: Gemini on a free tier could return an empty answer when rate-limited, so we changed the collector to stop rather than record a false absence. Answers that legitimately name nobody also vary by surface. In Houston, 59 of 492 Perplexity answers (12%) named nobody against 9 of 479 (1.9%) from Gemini through its API.
5. Who wrote the prompts? A prompt that contains your own name tests recall of a name you supplied. We would want brand prompts reported apart from category prompts, because the blend hides the number a stranger’s search actually produces.

Our own measurement is the product of those settings: a question set fixed before anything runs, multiplied by engines, by three runs, by clean sessions. In a guarantee re-test we ask each question three times per engine and count it when you are named in two of the three. A single daily reading on a dashboard is one cell of that grid, so we read the trend over weeks and ignore the day.

What do these products cost on the vendors’ own pages today?

Self-serve plans run from free to about $500 a month, and the useful comparison is price per prompt and engine, because the allowances differ more than the headline prices do. Every figure below was read on the vendor’s page on 7 November 2026.

The last column assumes each prompt runs on each included engine once a day, which the vendor states for Otterly, Peec, Semrush and Frase. It is our calculation, and it moves if your plan runs fewer engines than the allowance. Within one vendor the larger plan is no dearer per pair (Otterly $0.47 against $0.48, Peec $0.47 against $0.63), while the suite module costs about double the cheapest entry plan.

The other price shapes we found are harder to line up:

• ZipTie sells through a configurator, from $25.65 a month, with a free 7-day trial of 25 daily prompts on three engines.
• Ahrefs Brand Radar shows custom prompt tracking from $50 a month (the Lite, Standard and Advanced tiers cover 5, 10 and 20 prompts with no tier price printed) and its index from $199 a month. The page also says Prompt Tracking is free on paid Ahrefs plans and does not explain how the two fit together, so ask which applies to you.
• Rankscale prices in credits, from $17 a month on yearly billing, and prints answers instead of prompts.
• Scrunch lists $250 a month billed annually or $300 month to month for 350 custom prompts, and AthenaHQ lists $295 a month for its Starter plan with credits instead of a prompt cap.
• Profound and AirOps print no plan price.

Third-party prices drift fast. An earlier version of this post quoted ZipTie at $58.65 to $84.15 from Zapier’s February 2026 test. Zapier updated its page in September 2026 and now says from $35.63, while ZipTie’s own page says from $25.65. Zapier’s current list does not include Profound, and the only Profound price we could find is Siftly’s “about $499 a month”, which Siftly marks “indicative only” on a page last updated in June 2026. Profound’s own page prints no such price. Re-check on the vendor’s page the week you buy.

Who writes the “best AI visibility tools” lists, and do they sell one?

Most do. Of the nine pages we could read on 7 November 2026, six are published by a company selling an AI visibility product, a seventh is written by an employee of one of them, and an eighth agency offers its own dashboard. Five of the six sellers list their own product first, four of them as an explicit number one.

Two entries are murkier than a yes or no. Brainlabs describes a dashboard built for its clients, though its list does not rank anything. The last page publishes a scoring model with weightings, and says all inputs were scored with “Profound’s proprietary AEO ranking model”, with Profound on top. Its author is Profound’s Founding Marketing Engineer, per his own site, and he also wrote Profound’s own 18-tool list. The post’s only disclosure is that it represents his personal views and not those of Profound. We also tried a tenth page, a Search Engine Land guide, which returned an access error to us, so we left it out.

Naming the neutral pages matters as much as naming the sellers. Zapier prints a price for seven of the nine tools it lists and marks Similarweb and Yext Scout as by request, and it discloses that it is a customer of Petra Labs, one of the nine.

One more receipt, because the ranking above rests partly on vendor research. Profound’s citation-source study covers 11.84 billion citations from eight AI models collected between 16 April and 16 July 2026, and finds roughly 57% of citations globally point to brand-owned sites, with ChatGPT at 47%, the lowest among the major answer engines it names, and Gemini at 69%. Useful numbers, and also the numbers a competitor is being scored against.

Which type of AI visibility product should you buy?

Buy the type that matches how many prompts, markets and clients you have to report on, and buy nothing at all if you have never run a manual baseline. These are the cases we see most.

Our view is that most single-brand buyers should stop at row two. More prompts lower the cost per pair, and they leave the sampling problem from the previous sections exactly where it was.

Do you need one yet?

If you have never run a manual baseline, run it first. It takes an afternoon, it shows what a good prompt list looks like before you pay someone to run one, and it tells you what a free trial would.

1. Write 20 real buyer prompts: five category prompts (“best real estate agent in Tampa, FL”), five comparisons, five problems in the buyer’s words and five that name you.
2. Run each in four engines: ChatGPT, Perplexity, Google AI Overviews and Gemini, logged out, in a fresh chat each time.
3. Log two columns: brand named yes or no, and every URL the answer cited. The cited URLs are the useful half, since they show which pages the engines already trust in your category.
4. Repeat in 30 days. Now you have a trend, built by hand, for nothing.

That is 80 checks and about two to three hours. Our walkthrough on running an AI visibility audit yourself covers prompt writing in detail, and the clean-room protocol in our competitor tracking guide covers the session settings. A paid product starts earning its fee when you track more prompts, markets or brands than one person can hand-run each month, or when a client or board needs the report on a schedule.

Which companies offer AI citation tracking for real estate?

None of the pricing or product pages we read describes a real estate prompt set or a city-level view. Every product above is horizontal: it tracks whatever prompts you load.

That describes the job more than it faults the tools. For a brokerage the prompts are local, “best real estate agent in Tampa, FL” and not “best real estate agent”, and the answer changes by city, so the unit of tracking is the metro. It changes within a city too: in our 500-query Houston study, an estimated 82.6% of the agents, teams and brokerages named appeared in just one of 40 Houston areas.

Read the plans above through that lens:

• Otterly Lite’s 15 prompts cover one city across four engines and leave no room for a second city.
• A 25-prompt plan from Semrush or HubSpot fits one metro with a few comparison and brand prompts alongside.
• A tracker limited to one engine sees only part of the picture. In our 150-question real estate study, of 1,297 cited domains ChatGPT cited 305 (24%), Gemini 50 (4%) and Perplexity 618 (48%), and 83% were cited by only one of the five engines.
• Write the city and state into every prompt yourself, because a country setting will not do it for you.

What no dashboard hands a brokerage is the list of sites the engines cite instead of you, which is the outreach list. Our redacted sample AI visibility audit for a real estate firm shows the prompt set, the scorecard and that list for a firm that was named in none of 38 answers. To start the list before any audit, our open list of the domains AI cites on real estate gives per-engine counts for every site the five engines cited.

Does a dashboard actually fix anything?

No. It is an instrument, and the work that changes the score happens in the pages and signals underneath it. A thermometer is useful, and it has never lowered a fever.

In our June 2026 study of 181 US ecommerce brands, 87% were absent from all four AI engines when each engine was asked a buying question in the brand’s own category. Every one of those 158 could have bought a dashboard and watched the same zero, weekly, in a nicer font.

The work that moves it, in the order it usually pays off: crawler access so engines can fetch you at all, page structure that can be quoted without editing, schema that makes your entity unambiguous, and then the third-party mentions that sit off your domain and take longest. The GEO paper accepted to KDD 2024 reported up to 40% higher visibility in generative engine responses from content changes such as adding citations, quotations and statistics. That is a content result, and no tracker performs it for you. The diagnosis half is in what generative engine optimization is.

Our Bright Realty case study shows the search-side results of that groundwork for one brokerage (mainly Search Console clicks and impressions, plus a directional Semrush count of 35 cited pages), and our real estate SEO agency page covers how we run it. Buy the instrument if you need the instrument, and keep it separate from the treatment.

What happens to your contract if this category consolidates?

The market is consolidating while you shop in it, and buyers are signing annual deals into that. Two events from the last twelve months are worth knowing before you commit to a 12-month term.

Adobe agreed to acquire Semrush for about $1.9 billion, completed on 28 April 2026.

Adobe agreed to acquire Semrush for approximately $1.9 billion in an all-cash deal at $12 per share, announced on 19 November 2025, and Semrush’s newsroom dates the completion to 28 April 2026. Adobe framed the purchase around brand visibility in generative AI. Semrush’s AI toolkit page still reads $99 a month per domain, billed annually, so check how it is packaged under Adobe before you rely on today’s price.

Profound raised $96 million at a $1 billion valuation in a Series C led by Lightspeed Venture Partners, announced on 24 February 2026, taking total funding to $155 million, and it serves more than 700 enterprises. Money at that scale tends to buy smaller tools in the same list you are choosing from.

Three buying rules follow, and none cost anything:

• Prefer monthly billing over annual until the category settles.
• Get it in writing before you sign that you can export your historical prompt data if the product is absorbed, repriced or shut down.
• Re-check the price on the vendor’s own page the week you buy. The ZipTie example above shows three different prices from three sources.

---

Want the baseline without the afternoon? Our Free AI Visibility Audit takes your website and an email address, runs 50 buyer questions across ChatGPT, Perplexity, Gemini and Copilot, and sends back 200 screenshotted answers showing which ones name you and which name a competitor, within 48 hours. It is free, with no card and no call.