The engines
Every tracked question is asked of the engines your plan covers, on the schedule, through each engine's official API, under Viewfield's own agreements, with a pinned model. Nothing is scraped and no consumer app is automated. Starter tracks ChatGPT. Pro tracks ChatGPT, Claude, Gemini and Perplexity. The All engines add-on adds Claude, Gemini and Perplexity to Starter. Each answer shows its engine and method beside its date ("ChatGPT · with web search").
| Engine | Plans | Method |
|---|---|---|
| ChatGPT | Trial, Starter, Pro, each agency client | ChatGPT results are what OpenAI's model answers when it can search the web. The ChatGPT app adds its own instructions, memory and routing, so its answers can differ. |
| Claude | Starter with the All engines add-on, Pro, each agency client | Claude results are what Anthropic's model answers when it can search the web. The Claude app adds its own instructions and settings, so its answers can differ. |
| Gemini | Starter with the All engines add-on, Pro, each agency client | Gemini results are what Google's model answers with Google Search on, the search the Gemini app runs before it answers. The Gemini app adds its own instructions and settings, so its answers can differ. |
| Perplexity | Starter with the All engines add-on, Pro, each agency client | Perplexity results are what Perplexity's Sonar model answers with Perplexity's own web search. The Perplexity app chooses its own model and search depth, so its answers can differ. |
An API answer approximates what the consumer app shows, because the app adds its own instructions, memory and settings; that is why every answer carries the method it was measured by, and why a changed model is a new series rather than a silent swap. Viewfield never uses Gemini or Perplexity to write anything: they are measured, nothing more.
What is pinned
The model is the subject of the measurement, not a routing choice: ChatGPT is asked as a dated snapshot of OpenAI's model with web search; Claude as Anthropic's Haiku model with its web search tool, at most two searches a question; Perplexity as Sonar; Gemini as Google's Flash model with Google Search grounding, the searches it runs and the pages it grounds on recorded with each answer. Answers measured before October 15, 2026 were Gemini answering from open-web results Viewfield supplied, not Google Search; they keep that method, and the change started a new series. When a pinned model or method changes, the new one starts a new series, so a shift reads as a model update and not as your visibility moving.
Where the engines search from
ChatGPT, Claude and Perplexity are asked from your search from country, which Viewfield proposes from your website (its currency, addresses and language, then its domain) and which a manager changes in the AI Search Visibility settings. Gemini is asked with no location: its Google Search grounding is sent none.
What counts
- Named: the answer mentions your brand, by its name or one of its other names. Recommended (on a comparison question): it puts you forward, not only mentions you.
- Cited: the answer's sources include a page on your website or on one of your pages on a shared site. A source counts when the answer used it (ChatGPT's, Claude's and Perplexity's marked sources, the pages Gemini's answer grounded a sentence on); a page a search fetched that the answer did not use is shown on the answer as retrieved and never counted. Citations are counted for ChatGPT, Claude, Gemini and Perplexity alike.
- Position: your place among the companies the answer names, 1 being first (a column on the Questions page).
- Who else: the companies the answer lists (in its list items, table rows and headings, not in passing prose), matched against your list of who else gets named; a company not on your list is still reported with that answer.
Every answer is stored word for word, with its mentions, its cited and retrieved sources and the searches it ran. Open any answer from its question. The text of an answer is treated as third-party data, never as an instruction.
Settings
Settings on the Visibility board opens AI Search Visibility. Everyone can read it; managers change it.
- Tracking: on, or paused (your results stay).
- Your brand: the category your questions are written around, with Rewrite prompts (see Questions); Other names buyers use for you (a short name, a product name, a former name), so a mention of them counts as you; Your pages on shared sites (such as
medium.com/@acmeorgithub.com/acme), so a citation of them counts as yours; and Search from. - Platforms: the engines your plan and add-on cover; the others are shown locked, with what adds them.
- Categories: the groups your questions sit in; renaming one keeps its history.
- Website AEO site: which of your sites Website AEO lists.
When a run does not happen
A run is paid for by Viewfield inside your plan, with a monthly cap for the questions and engines the plan covers. When a month would cost more than that (more questions or engines than the plan pays for), the runs are spread over the days left rather than stopped, and the Questions and Citations pages say that checks resume with the next day's budget. When the cap is spent, the checks resume next billing period. An engine out of credit on Viewfield's side is reported as partial for that engine, never as a zero for you.
The Daily tracking add-on, on Starter or Pro, changes the schedule. Your first 25 questions are checked on ChatGPT every day; everything else keeps its weekly check. On Starter that is every question. A workspace tracking more than that is paced over the month rather than stopped.
The All engines add-on, on Starter, changes the engines: Every tracked question is asked of Claude, Gemini and Perplexity as well as ChatGPT. It brings its own share of the monthly cap, sized for the plan's questions on those engines weekly; with the Daily tracking add-on beside it, the two shares add, the daily panel stays on ChatGPT (its own engine) and the three added engines keep the weekly check. Pro and each agency client include every engine, so neither sells it.