Last updated October 2026.
Ask a vendor how its tool reads ChatGPT, and you will usually get a features list, not an answer. Prompt count. Engine count. Refresh rate. The plumbing behind the number, the thing that decides whether a reading is even physically possible to collect, almost never makes it onto a pricing page.
That plumbing comes down to two core methods, plus one genuinely different third option among the tools in this comparison. The method a vendor uses is not a branding choice. Two of the five engines this lab covers leave no other door in, and the third method exists precisely because one vendor decided the standard two weren’t the real problem to solve.
This matters past curiosity. If you buy a tracker to guide content or PR decisions, the collection method sets the ceiling on what that tracker can ever show you. A tool locked into API access cannot see what a real ChatGPT.com visitor sees, no matter how clean its dashboard looks.
Two lanes, and a third that isn’t either one
UI automation points a headless browser at a real chat product, ChatGPT.com or a live Google search, and reads what loads on the page. It sees exactly what a person sees: the answer, the citations, and the live web browsing behind both. It is also slow to run at scale and an easy target for rate limits and layout changes.
Direct API access skips the browser. A developer key calls a model’s own endpoint and gets a structured response back. It runs faster and does not fight anti-bot defenses, but the answer an API returns is not guaranteed to match what the identical prompt shows inside the consumer app. An API call also has no browsing history, no account, and no location to draw on, so it cannot see the personalized version of an answer a logged-in user gets.
Ahrefs Brand Radar builds a third path around a different question entirely. Not how you query an engine, but which prompts you send it. Instead of a panel an analyst writes by hand, it starts from real search demand and runs that list through every platform it tracks. More on that below.
Why ChatGPT, AI Overviews, and Copilot force the UI lane
OpenAI publishes a developer API, and it is genuinely useful. It is not the same product as ChatGPT.com. The consumer app runs live web browsing and drops source links inside the answer. A standard API call does not reproduce that behavior by default, and nothing OpenAI publishes lets a third party pull the exact citation list a ChatGPT.com visitor just saw.
Google AI Overviews sit in the same spot. There is no AI Overviews API, public or paid, from Google. SerpApi, a search-data provider built for exactly this problem, confirms the only way to capture an AI Overview and its cited sources is to render an actual Google results page and parse what comes back.
Microsoft’s consumer Copilot sits in the same category. Microsoft publishes developer tools for Copilot Studio, Security Copilot, and Azure, but none of them reproduces the live, cited answers the consumer product at copilot.microsoft.com shows a visitor. A tool tracking that surface reads the same page a person would.
Three of the five engines in this comparison, then, are not a choice at all. Any tool that tracks ChatGPT, Google AI Overviews, or Copilot has one door in: UI automation, run directly or built into a prompt-database method like Ahrefs Brand Radar’s.
Where a direct API actually works
Gemini and Perplexity both leave the other door open. Google publishes a public Gemini API for developers, and Perplexity’s API returns built-in citations alongside its answers, no browser required. A vendor tracking either engine can call the endpoint directly and skip the automation problem entirely, at least for that one surface.
Evertune is the clearest example of a tool built around this option. Its published methodology runs a dual-layer approach: querying foundation models directly through their APIs, and separately capturing what the consumer apps answer to the same prompts. Comparing the two layers shows whether a visibility gap sits inside a model’s own training or inside the retrieval layer a chat product bolts on top of it, a distinction a single-lane tool cannot draw.
That gap is also the catch. An API call can hit a different model checkpoint than the one live in the app that week, or skip a browsing tool the consumer product switches on by default. Neither reading is wrong. They measure related but different surfaces, and a fair comparison needs to know which one produced the number in front of you.
Ahrefs Brand Radar’s third method
Ahrefs Brand Radar does not compete on which lane queries an engine. It competes on which prompts get sent down that lane in the first place.
Its own documentation describes the process directly: it takes real queries from its keyword database, expands them into natural questions, and runs the resulting 471 million-plus prompts through each AI platform it tracks, ChatGPT, Perplexity, Gemini, Copilot, and AI Overviews included (Ahrefs, 2026). That is a fundamentally different starting point from a panel an analyst configures by hand, and it is the reason this comparison treats prompt-database sourcing as its own category rather than folding it into API access.
Which lane feeds which engine, tool by tool
Here is what five tools in this comparison do, based on what each one discloses and on which engines leave no alternative to UI automation in the first place.
| Tool | ChatGPT | Perplexity | Google AI Overviews | Gemini | Microsoft Copilot |
|---|---|---|---|---|---|
| Profound | UI (forced) | Not disclosed | UI (forced) | Not disclosed | UI (forced) |
| Peec AI | UI (forced) | Not disclosed | UI (forced) | Not disclosed | UI (forced) |
| Ahrefs Brand Radar | Prompt-DB | Prompt-DB | Prompt-DB | Prompt-DB | Prompt-DB |
| Evertune | API + UI | API + UI | API + UI | API + UI | API + UI |
| Temso | UI | UI | UI | UI | UI |
UI means UI automation. API means a direct call to a model endpoint. Prompt-DB means Ahrefs Brand Radar’s real-user-prompt database method. “Forced” marks an engine with no public citation API at all, so UI automation is the only physically possible route, whether or not the vendor says so. Plain labels reflect what the vendor states about its own method. Temso states directly that it reads real user interfaces rather than model APIs, across every engine it tracks.
What this means before you trust a number
A visibility percentage is only comparable to another percentage collected the same way. A UI-automation reading of ChatGPT and an API reading of Gemini measure two different products by two different methods, and stacking both on one chart hides that difference instead of resolving it.
Before you trust a citation number, ask the vendor three things: which lane produced it, for which engine specifically, and whether that lane changes engine by engine inside the same platform. A vendor that can answer all three in plain language is telling you something most pricing pages never do.
Check that answer against the full field on the LLM visibility benchmark, where every tracker on this site is graded against the same testing protocol.