AI BEACON · PROMPTS
The questions your category asks,
and where you stand on each one.
A keyword tells you what someone typed. A prompt is the whole question, and the answer to it either names you or it does not. Prompt coverage is the set of questions you have chosen to watch, and your standing on every one.
Four hundred keywords tracked. Not one prompt.
The gap tends to show up in the same three conversations.
Somebody asks what the company ranks for in ChatGPT. There is no report for that, so the answer becomes an anecdote about a search someone ran last week.
The prompt list was drawn up when the account started and has not been looked at since the product line changed.
Visibility is being tracked by pasting answers into a document, which works until anyone asks what changed since last month.
Pick the questions, then watch the answers.
Coverage is a decision before it is a measurement. The prompts you choose are the ones you will be judged against.
Add your prompts
The questions a buyer would type, in their words. Keyword Explorer is where you go looking for them.
Research questions and keywords →Sort them by market and stage
A comparison question and a how-does-it-work question need different pages behind them.
Each engine answers in its own right
The same prompt produces different answers and different sources on each engine, so each is run separately.
Read the movement
The same prompts run again over time, so a change in presence is recorded rather than recalled.
SEORCE has observed 441.4M AI prompts: 33.9M from ChatGPT and 407.5M from Google AI Overviews, as of October 2026. That corpus covers those two engines only.
Prompts, grouped, and scored where it counts.
Four things to work with once the list is in.
Prompts you chose, not prompts we guessed.
Coverage is only useful if the questions are the ones your buyers ask. You write them.
- Your own wordingAdd prompts as a buyer would phrase them, including the clumsy ones.
- By marketEach site or market is its own project, so nothing gets averaged across them.
- Edited as you learnA prompt list is a living thing. Swap one out when the question stops mattering.
Named in the answer, or absent from it.
Whether a model names your brand inside the answer is the signal that matters most, because that is the part a reader sees.
- Per promptA clear reading on each question you track.
- Per engineSeven engines plus Google AI Overview, AI Mode and Search, read separately.
- Beside the sourcesA mention and the page behind it sit next to each other.
The same question, asked again.
One reading is a snapshot. The value is in the second and the tenth, where a slide shows up while it is still worth doing something about.
- A record per promptPresence over time, rather than a memory of a good week.
- Sentiment alongsideA score from 0 to 100 for how the answer described you.
- Changes worth a lookA prompt you held last month and lost this month.
A missing prompt is a piece of work.
The list of questions you are absent from is the most useful output here, because each one points at something to write or to fix.
- Absent and worth winningThe prompts where a buyer decides and your brand is not in the room.
- Handed to Content StudioA gap becomes a brief and a page in your brand voice.
- Handed to Technical SuiteWhen the page exists and cannot be read, the problem is technical.
Keywords tell you about traffic. Prompts tell you about answers.
Both are worth tracking. They answer different questions, and only one of them tells you what a model says when a buyer asks for a recommendation.
A typical stack
- –Keyword positions, with the answer box ignored
- –Visibility judged from the odd screenshot
- –One engine checked, usually the popular one
- –No record of the questions you have never appeared for
With prompt coverage
- ✓Prompts you chose, read on every engine
- ✓Presence recorded per prompt and per engine
- ✓Seven engines plus Google AI Overview, AI Mode and Search
- ✓A standing list of the questions you are absent from
The rest of AI Beacon, and where this one sits.
Coverage is the starting point. These are the pages that tell you what to do with what it shows.
What is prompt coverage?
Prompt coverage is the set of AI prompts a brand has chosen to track, and its standing on each one. In AI Beacon you add the questions your buyers ask, group them by market and by stage of the decision, and each prompt is run across 7 AI engines plus Google AI Overview, AI Mode and Search. For every prompt you get whether your brand was named in the answer, how it was described on a score from 0 to 100, and which sources the answer drew on.
Key facts
- You choose the prompts; they are not generated for you
- Each prompt is run on every engine separately, because answers and sources differ by engine
- Presence in the answer is the primary signal, and citations are the second-order signal
- Each site or market is its own project
- AI engines publish no prompt volume, so coverage counts prompts rather than searches
- SEORCE has observed 441.4M AI prompts (33.9M ChatGPT, 407.5M Google AI Overviews, as of October 2026)
Last reviewed 15 October 2026 by the SEORCE team.
Before you pick the first ten.
Ten is enough to learn something. Pick the questions you would most want to be recommended in, write them the way a buyer would, and add more once you can see how the first set behaves.
No. The AI engines publish no prompt volume, so nobody can give you a reliable search count for a prompt. Coverage measures your standing on the questions you track, not how many people asked them.
Keyword Explorer is built for that: question keywords, People Also Ask research and intent classification across 87M+ unique keywords, growing every day. Sales calls and support tickets are the other good source.
No. A keyword is what someone types into a search box. A prompt is the whole question put to a model, and it usually carries context a keyword does not, such as a budget, a team size or a tool they already use.
Yes. The same prompt produces different answers and different sources on each engine, so running them together would hide exactly the differences you need.
Open it in prompt diagnosis to see what the answer was built from. If no page of yours covers the question, Content Studio writes one. If a page exists and cannot be read properly, that is a job for Technical Suite.
Pick ten questions. See where you stand.
Add a site, write the prompts your buyers would ask, and read what comes back from each engine.