Skip to main content
Search increasingly happens inside an answer, not a results page. AI visibility measures the part of that you can actually observe: for the questions your customers ask, do the answer engines name your brand, and do they read your pages?

​
How it works

  1. Questions. daoco writes a dozen or so customer-style questions from your brand context: your audience, the problems you solve, your category, and your competitors. Most of them do not name your brand, because a question that names you only proves you can be looked up.
  2. A run. Each question goes to every answer engine daoco can reach, and the answers are parsed for your brand, your cited pages, your position among the brands named, and which competitors appeared.
  3. A scorecard. Rates per engine, the per-question grid, competitor movement since your last run, and the fixes the results support.

​
Where it lives

Open Home, choose Website, and open the AI visibility tab. That is the durable home for this work and it survives refreshes and new chats. The hero band shows your mention rate, how often your own site was cited, and how many questions at least one engine named you in, with a chip for each engine. Check my AI visibility runs it the first time and Check again afterwards. In chat, type /ai-check to run the same check. To repeat the check every week, turn on your assistant’s AI visibility focus area. See Keeping it running. For a personal brand the tab is called Found by AI. The questions are the ones a recruiter, hiring manager, client, or event organizer would ask (for example, “recommend a speaker on design systems”), and the check records whether the answers name you. If your goals are creator goals (growing an audience, brand deals, your own product), the questions are the ones a follower, brand partner, or buyer would ask instead, such as “best creators to follow about budget travel”, “which creators should an outdoor brand partner with”, or, if you sell your own product, “best course on sourdough for beginners”. With both career and creator goals, the set is split between the two. Change your goals in Brand Hub and the next generated set follows them. If your site lives on a shared host such as Substack, Medium, Linktree, or GitHub Pages, only your own pages there count as cited. A link to someone else’s page on the same host does not. The Questions asked card is where you manage the question set: type a question and choose Add, or hover a question and remove it. Removing a question stops asking it and keeps its history. Hover a question to see what each engine said and which sites the answers were sourced from. Edits apply on the next run. A chat run also opens an AI visibility panel with the same scorecard.

​
Ask for it in chat

​
The engines

ChatGPT and Gemini are asked the bare question, as a person would type it, from your market. daoco takes the market from the location and language your tracked ranking keywords use most, and uses the United States in English when you track none. If the app cannot be reached for a question, daoco asks that engine’s API instead (OpenAI with web search, or Gemini with Google Search grounding). Those answers are labelled API approximation on the scorecard, because an API is close to the app but is not the app. An engine only runs where daoco can reach it. Engines it cannot reach are reported as not asked and are never counted as a zero, so a rate you see is always over answers that actually came back.

​
New baseline after the engine change

Checks before October 2026 asked ChatGPT and Gemini through their APIs. The first check after the switch to the apps’ own answers is a new baseline: it is not compared with earlier checks, so it never shows gains or losses that only reflect the change of engine. Later checks compare with it as usual.

​
Reading the scorecard

Mentioned means an answer named your brand. Cited means the answer used a page on your own domain as a source. They are different problems:
  • Mentioned but not cited is the common case. Engines know you from third-party sources and are not reading your site. That is a citability problem on your pages.
  • Absent while competitors are named is the strongest signal you can get, and the fixes based on it are the ones worth doing first.
  • Absent with nobody named usually means the question was not a recommendation question. Replace it.
Position is where your brand first appeared among all the brands an answer named. A single run is a sample. Answer engines are not deterministic, and the same question can return a different set of brands later. Movement only means something against a previous run.

​
Fixes

The Fix action follows the same rule as SEO fixes. It needs a usable CMS or GitHub connection before it starts. daoco inspects the connected resource and prepares a reviewable change. If the target is outside that connection’s reach, the run stops with that blocker instead of returning a generic plan. Each run joins its results with what daoco already measures per page (FAQ sections, definition blocks, comparison tables, author attribution, original data, schema markup, and freshness) and lists the gaps on pages that should have answered the questions you lost. daoco also notes who publishes each source an answer cited. When engines answered a question from your competitors’ own sites and blogs, and no page of yours was cited, the fix is a new page on your site that answers that question. These fixes come first. A review site, directory, community, or independent publication that engines keep citing is listed as a place to get your brand in front of them. Vendor sites and sources whose publisher could not be identified do not produce presence recommendations. Every fix names the page or source it applies to and the measured signal it came from. Nothing is suggested that the run did not observe. When a check completes, its fixes are recorded alongside your SEO findings, so a Fix run reads the recorded finding rather than a summary of it, and the same gap measured again refreshes that finding. If recording failed, the Fix run has only the fix’s title to work from and says the finding was not recorded. Without a usable connection the control reads Connect CMS or GitHub and takes you to Connections instead.

​
Keeping it running

Turn on your assistant’s AI visibility focus area to run the check every week. Its Weekly AI visibility review runs a fresh check every Thursday, reports whether the pages written after earlier reviews are now cited, then writes and saves a page for the biggest question where engines name a competitor instead of you. See focus areas. It never publishes anything; changes to your site wait for your approval.

Pause or change it

Switch the weekly review off under Assistants. Add, edit, re-run, or remove individual questions from the Questions asked card or from chat, or ask daoco to regenerate the whole set. Your own questions are kept when it does.

​
Limits

  • A single run asks at most 20 questions, whatever your plan.
  • Starter workspaces keep 10 saved questions; Growth keeps 30.
  • Each question costs one answer per engine: a DataForSEO task for ChatGPT and Gemini, and a model call for Perplexity or any API fallback, including its search fees. A further model call reads each successful answer to identify the brands and source publishers. All of it is billed like other daoco usage.
  • An app answer can take up to about two minutes, so a full check can take several minutes.