AEO / GEO Prompt Universe

Every tracked prompt runs on four engines:

• ChatGPT, with web search on, so answers reflect current sources.
• Claude.
• Gemini.
• Google AI Overviews, the summary Google shows above search results. Some searches show no overview, and that is counted too.

You pick a daily or weekly refresh. When you track a prompt you can also limit it to the engines it fits best.

1. 1

   Read your docs and conversations

   EarShots reads your docs and relevant posts with confirmed publication dates. Archived, noisy and estimated-date posts stay out. New prompts are added as your product changes.
2. 2

   Find prompts in the Prompt Planner

   Search a topic, or paste your own list, to find prompts you do not track yet. Each prompt gets a score from 1 to 100 and a tag for the engines it fits best. Similar prompts are grouped into themes, up to 10 per search.
3. 3

   Ask the engines

   Tracked prompts run on ChatGPT, Claude, Gemini and Google AI Overviews on their cadence. EarShots stores each answer and the sources it cites.
4. 4

   Check every answer

   A decision model reads each answer and decides whether it really mentions your product or a namesake. Unknown brands in answers are checked against the web before they count.
5. 5

   Review and act

   See mentions, citations and position per engine over time, the sources engines cite instead of you, and get recommendations for any prompt.

Every search in the Prompt Planner is saved as its own bucket. Opening a bucket shows only the prompts that search found, and the prompt score and trend line above the table are worked out from that bucket alone.

• Seven active buckets. A workspace can keep up to 7 active buckets. To run an eighth search, archive one first.
• Archiving keeps the prompts. An archived bucket keeps its prompts and can be restored when a slot is free.
• All buckets. Turn on All buckets to see every prompt from every bucket in one table.

• Mentioned in: the share of answers that mention your product.
• Cited in: the share of answers that link your site as a source. Add a docs URL to track this.
• Average position: where you are mentioned when you are mentioned, first, second and so on.
• Brand ranking and share of voice: how often you are mentioned next to the competitors you track.
• Top sources: the domains engines cite, with the share of your prompt answers each one is cited in. Domains are labeled as you, a competitor, or a source type such as docs, forums or editorial. Competitors come from your Knowledge Hub first.

Filter by engine or by a single prompt. Look back 7, 30, 90, 180 or 365 days, or pick a custom range of up to one year.

Counting a brand name with text search gets it wrong often. A product name can also be a person, a place, or another tool. The checker is a decision model that returns a typed yes or no with a probability, not generated text.

• Is it you? The answer is read next to your Knowledge Hub, including namesake rules, to decide whether the mention is your product.
• Is it a real brand? A new name seen in answers is searched on the web to decide if it is a product or company. Real ones join share of voice. Words like "Verify" or "Click" are dropped.
• Unsure cases go to people. Readings close to 50 percent land in a review queue. Labels from that queue measure accuracy over time.

Open any tracked prompt to see What to do. EarShots reads the captured AI answers, the brands and domains engines name and cite, and your docs, then lists recommendations, each with the evidence behind it, concrete steps, impact and effort. EarShots does not draft docs pages. It drafts replies on X, LinkedIn and Reddit for your approval.

Starter tracks 100 prompts. Growth tracks 250. Enterprise tracks prompts on credits, sized with the EarShots team for your volume. A workspace without a paid plan can open the read-only preview but cannot track prompts.