When a shopper asks ChatGPT for the best sunglasses for driving, somebody's store gets recommended. AI answers is where you find out whether it's yours.
Open it at /dashboard/answers.
The idea
AgentReady writes a set of tracked questions the way real shoppers actually ask them, grounded in what your store sells. Then it asks the real assistants — ChatGPT, Gemini, and Perplexity — and records the answers verbatim. Every claim on the page comes from a counted, stored answer. Nothing is a score we invented.
Each question sits on a four-step ladder:
- Invisible. The assistants recommended other stores, not yours, in every answer we collected.
- Sometimes. You appear in some answers but not most.
- Recommended. You are a regular answer.
- Cited. An assistant linked to your own page as its source. This is the top of the ladder.
Movement between those steps is the whole point. The section's centerpiece shows each question's position and, once there is more than one measurement, the path it took.
Tracked questions
The recommended questions come from your actual catalog and the searches your shoppers actually make, phrased with purchase intent so a store like yours can genuinely win them. Each one shows the store phrase it was grounded in.
You can add your own. The form checks each phrase honestly: a very broad question like "best sunglasses" is accepted but labeled a stretch goal, because today only household names win it. A how-to question is flagged too, since assistants answer those with instructions rather than store recommendations.
Open any question to read the real answers, engine by engine, including which competitors were named and which sources the assistant cited.
Honest numbers, by design
AI assistants vary their answers run to run. So the page reports counts, never certainties: "named in 3 of 8 answers," not a rank. A question is only called Invisible after enough runs across at least two different engines. One measurement establishes a baseline; movement only appears once there is something real to compare.
Actions: from gap to fix
Under the movement view, AgentReady turns your sharpest gaps into cards, and every card is explicit about who does the work.
- We can do this. One press and AgentReady drafts the page built to become the cited answer, or a grounded comparison against the competitor the assistants keep naming. It lands in your Reach review queue; nothing publishes unreviewed.
- You do this. Some changes live in places only you control. The card says exactly what to change, and "Mark as done" starts the clock.
Suggestions never contradict themselves: a declined card stays gone for a month, an applied one is never re-pitched, and AgentReady won't suggest writing a page you already have.
Then it watches
Every applied action is re-measured at 7, 14, and 28 days. Only the affected questions get re-asked, and the verdict is worded the way honest measurement demands: "moved since," "no change yet," or "too early to tell." AI answers vary run to run, so movement is always a comparison of measurements, never a single answer.
Availability
AI answers is included on every paid AgentReady plan. Free stores get a real taste: one full measurement of their generated questions, with the results readable forever. Tracking your own phrases, the action cards, and re-measurement are part of the paid plans.