AI Visibility Tools: How to Choose Among 176 Trackers

AI visibility tools send a fixed list of prompts to ChatGPT, Perplexity, Gemini and Google's AI Overviews or AI Mode on a schedule, then report whether your brand was named, which pages were cited, and which competitors appeared instead. They measure; generative engine optimization tools are the ones that change what the engines can read. We count 176 of them in the INITE Atlas, with a median entry price of $49 a month, so finding one is easy and choosing is the work. Choose on answers per month (prompts times engines times runs), on whether the tool shows the cited URLs, and on whether it covers the engines your buyers use. For one brand, Otterly.AI, Peec AI, Trakkr and Rankscale cover the common case at $29 to $100 a month. Profound and Scrunch are built for larger teams, and Semrush and Ahrefs make sense if you already pay for their SEO suites.

What do the ten established tools cost and cover?

Pricing pages read 2026-10-07. Prices are monthly billing unless marked; "engines" means what the entry plan includes without add-ons.

Why does the count of 176 matter?

Because it changes what you are shopping for. Of the 176 trackers in the Atlas (read 2026-10-07), pricing could be read for 166; 124 of those have a free tier, and only 4 sell through sales calls alone. Of the 113 with a monthly price in US dollars, the median entry plan is $49, a quarter cost $29 or less, a quarter cost $99 or more, and the cheapest is $3.

So price at entry tells you little. Building a basic tracker means calling a few model APIs on a schedule and storing the answers, which is why so many exist. What separates them is how many answers they actually collect per month, which engines they query (through the consumer interface or through an API, which can answer differently), and what they show you about the sources behind each answer.

Which criteria actually separate them?

Six, in the order they usually decide the purchase:

1. Engines covered. ChatGPT is on every list. After that, coverage of Google AI Overviews, AI Mode, Perplexity, Gemini and Copilot varies, and several tools sell engines as add-ons. Track the engines your buyers use. For most B2B buyers that is ChatGPT plus Google; consumer categories often need Perplexity and Gemini too.
2. Answers per month. Prompts times engines times runs. This is the real unit of what you buy, and only some vendors print it.
3. Cited source URLs. A mention count tells you where you stand. The list of pages the engine cited tells you what to change, because those are the pages you need to be on or to outrank. Otterly, Peec, Trakkr, Rankscale, llmrefs, Scrunch and Ahrefs list citation or source analysis on their pricing or product pages; for the others, check that the entry plan includes it.
4. Competitor share of voice. How often you appear against named competitors on the same prompts. Scrunch caps this at 5 competitors on its entry plan; most others do not print a cap.
5. Export and API. CSV export is common at entry. API access usually is not: Peec, Profound and Scrunch reserve it for enterprise plans, Otterly starts it at $189 a month, Trakkr at $500, Rankscale at $229. llmrefs includes it at $79.
6. Price at entry, with its unit. "From $29" means 15 prompts at Otterly and 500 keywords at llmrefs. Always convert to answers per month before comparing.

What does one tracked answer cost?

Where the pricing page states the volume, the arithmetic is simple (pricing pages read 2026-10-07):

Otterly Lite works out near $0.016 if each of the 15 prompts runs daily on each of the 4 included engines, but its page does not state the answer count, so treat that as an estimate. Semrush and llmrefs do not publish an answer count either.

Cost per answer matters because of the next section: you need many answers per prompt before a number means anything.

What can't these tools tell you?

The biggest limit is statistical. Ask ChatGPT the same question twice and you can get two different lists of recommended suppliers, cited from different pages. The answer also depends on location, on whether the engine searched the web for that run, and on the model version that served it. Every tracker is sampling a moving distribution.

Even for market-leading brands, a single prompt can produce different recommendations day to day…

• quiqueqs, who builds an AI-visibility tool, on Hacker News (link), 6 February 2026

Most entry plans take one sample per prompt per engine per run. Ahrefs says so outright (one check per prompt, platform, location and update). Trakkr defines a data point as a single query to one model. Scrunch's 5,000 responses across 125 prompts and 4 engines come to 10 answers per prompt per engine per month, which is fewer than daily. So a single day's result for a single prompt is one draw, and a drop from "mentioned" to "not mentioned" overnight is usually noise.

Three practical consequences:

• Ask the vendor how many runs per prompt each figure is based on. If the answer is one, read weekly or monthly aggregates and ignore daily swings. llmrefs says it aggregates responses for statistical significance; ask any vendor that claims this what the threshold is.
• Track more prompts rather than checking more often. Thirty prompts run daily give a steadier monthly figure than five prompts run hourly, because the variation between prompts is part of what you are measuring.
• Do not compare scores across tools. Each uses its own prompt set, engines and formula. A visibility score of 40 in one tool and 25 in another can describe the same brand on the same day.

The other limits are about scope. A tracker reports what the engines said; it does not tell you why. Whether a crawler can reach your pages is a separate question, covered in which AI crawlers to allow. Whether being cited produces visits or revenue needs your own analytics. And the prompt list is your guess about what buyers ask, so the tool measures your guess.

We wrote about the gap between what a site does technically and how often it is actually named in our own AI visibility, measured: a site can score well on readiness and still be named rarely, and a tracker shows you that gap without closing it.

Which tool fits which buyer?

Otterly.AI is for someone learning the category. At $29 a month the Lite plan tracks 15 prompts daily on ChatGPT, AI Overviews, Perplexity and Copilot, with Claude, AI Mode and Gemini sold as add-ons. It is not for anyone who needs a trend they can defend: 15 prompts is a thin sample, and the next tier jumps to $189 for 100 prompts. The honest weakness is that the useful plan costs more than six times the advertised one.

Peec AI is for a single brand with a small marketing team that wants clean reporting. Starter is $95 a month for 50 prompts on any 3 of 6 engines, daily, with unlimited users. Extra models cost about $30 a month each on annual billing at Starter level. It is not for teams that need an API, which sits on the custom Enterprise plan. The weakness is the three-engine cap: covering ChatGPT, AI Overviews, AI Mode and Perplexity already means paying for a fourth.

Trakkr is for one brand that wants every major engine without add-ons. Growth is $100 a month for 50 prompts across 8 models daily, with 3 seats and one year of history. It is not for agencies on a budget, since the next plan (10 brands, REST API) is $500. The trial needs a card.

Rankscale (Atlas card) is for someone who wants to set the cadence and engine mix per prompt, from hourly to monthly. Pro is $99 a month for 1,200 credits. It is not for buyers who want a fixed bill: credits run down faster if you add Claude or hourly runs, and the REST API starts on Pro+ at $229.

llmrefs (Atlas card) is for SEO teams and agencies who think in keywords. $79 a month buys 500 prompts, all engines, unlimited projects and seats, CSV and API. It is not for anyone who needs daily data: reports refresh weekly. It also generates prompts from your keywords rather than letting you write each one, which suits broad coverage and fits a narrow niche less well.

Writesonic (Atlas card) is for teams that want tracking and an AI writing tool in one subscription. Starter is $79 a month billed annually for 50 prompts on ChatGPT, Gemini and AI Overviews, plus 15 AI articles. It is not for someone who only wants measurement: per answer it is among the more expensive, and Perplexity and the other engines are on Enterprise only.

Semrush AI Visibility Toolkit is for existing Semrush customers. $99 a month adds 25 tracked prompts, one domain for brand performance reports, prompt research and 10 CSV exports a day. There is no free trial on the toolkit, and its documentation names only ChatGPT and Google AI Mode for prompt tracking, so confirm the engine list before buying. Semrush One Starter, at $199 a month, bundles SEO with 50 daily prompts.

Ahrefs Brand Radar is for existing Ahrefs customers and anyone researching a market rather than one brand. Paid Ahrefs plans include a few tracked AI prompts (5 on Lite), custom prompt packages start at $50 a month for 2,500 checks, and the AI Visibility Index, built from prompts modelled on real searches, starts at $199. It is not the cheapest way to track Claude, which uses 8 checks per update.

Scrunch is for mid-size marketing teams that want monitoring, site audits and agent-traffic data together. Core is $250 a month for 125 prompts and 5,000 responses on four engines, 5 seats and 5 competitors. It is not for small budgets, and Gemini, AI Mode and Claude are on Enterprise only.

Profound is for enterprise brands with a dedicated team. The free trial runs 50 prompts daily for 7 days on ChatGPT, Gemini and AI Overviews with no exports and no prompt editing; everything else, including up to 9 engines, CSV and JSON export and API, is on a custom Enterprise contract. It is not for anyone who wants to see a price before a sales call.

How should you choose?

Start with what the engines say about you today, before paying for a schedule. The free AI visibility check on our site reads one site and shows how assistants currently describe it; a vendor's free trial does the same over a week. That baseline tells you whether you are absent, mentioned without a link, or cited, and each of those calls for different work, set out in how to get cited by AI assistants.

Then write the prompt list before you pick the tool: 30 to 100 questions a buyer asks before choosing a supplier, in their words, with competitors named in some. The list decides the plan size, and the plan size decides the price far more than the vendor does.