Run prompts across AI engines
Pick engines, run, and read mention, position, sentiment and cited URLs.
In shortTo run a prompt, open AI search → Brand visibility → Prompt research, click Run on the prompt, tick the engines in Select Engines to Run and click Run Now. When it finishes, Run Results shows, per engine, whether you were Mentioned, your Position, the Sentiment, the competitors named and the cited URLs. Each engine answer is charged separately, and failed answers are free.
When to run
Run a prompt when you want a fresh reading: to set a baseline, after you publish or update a page that targets the question, or before a client report. Because answers vary from one run to the next, a steady rhythm you choose yourself, such as the same prompts every two weeks, gives you a cleaner trend than many runs in one afternoon.
Before you start
- Have at least one prompt saved for the selected domain. See Add and research prompts.
- Work out the cost: prompts × engines × the per-engine price in the cost table below. The run window doesn't show a total.
- Check your balance. If it doesn't cover every engine you tick, the whole run is refused and you'll just see Run failed. Please try again.
Run a prompt, step by step
- Find the promptOn Prompt research, use Search prompts... if the list is long. Each row shows the prompt's Category, its Engines, Last Checked and Total Checks.
- Click RunClick Run at the right of the row. A window titled Select Engines to Run opens with all six engines ticked: ChatGPT, Claude, Gemini, Perplexity, Google AI Overview and Google AI Mode.
- Untick what you don't needEach ticked engine is a separate charge. If your buyers mostly use ChatGPT and Perplexity, run those. Untick the rest before you continue.
- Click Run Now and waitThe engines are asked one after another, so a six-engine run can take a minute or more. Keep the page open until the results window appears. If you leave, the engines that finish are still saved and charged, and you'll find them on the overview.
- Read Run ResultsA Run Results window opens with a count of successful and failed answers and one tab per engine.
Which engines to tick
You don't need every engine every time. Each engine finds and ranks sources in its own way, so your results will differ between them, and each one is a separate charge.
| Engine | Good to include when |
|---|---|
| ChatGPT | A sensible default for most prompts |
| Perplexity | Your buyers research before they buy and you want answers built on web sources |
| Gemini and Claude | You want a broad picture across assistants, or a client asks about them |
| Google AI Overview | The question is also a common Google search. Here it uses the default market, United States in English |
| Google AI Mode | You want Google's conversational answer for the same question, same default market |
Whatever you choose, keep the same engines from one round of runs to the next, so your trend compares like with like. If you need Google's answer for a particular country, run the separate AI Overview or AI Mode scan, which takes a country and language.
If an engine doesn't answer in time or returns an error, its tab in Run Results shows it as failed with no charge. Run that engine again later on its own: untick the others in Select Engines to Run so you only pay for the one you're retrying.
How to read a result
| Field | What it means | What to do with it |
|---|---|---|
| Mentioned / Not mentioned | Whether the answer named or cited your domain | Not mentioned on an unbranded buyer question is the gap to work on |
| Position | Where you came among the brands and domains in the answer, #1 being first | #1 or #2 usually means a top pick; #5 and lower is a long list |
| Sentiment | Positive, neutral or negative tone about you, or not mentioned | Neutral is common. Read the context line when it's negative |
| Context | One sentence on how you were described | Check it's accurate. Wrong facts point to pages to fix |
| Competitors | Other brands the engine named, with their positions | Who wins this question instead of you |
| Cited URLs | The pages the engine used as sources | The pages to beat, or to get your brand onto |
| Raw Response | The engine's full answer | Read it when the summary surprises you |
Back on the prompt row, Last Checked and Total Checks update. The clock icon opens Run History, every past result for that prompt. The checklist icon opens the prompt's Action plan, covered in the last lesson.
Engines don't give the same answer every time. Don't celebrate or panic over one run. Look for patterns across several prompts, engines and dates on the Visibility overview.
This page runs one prompt at a time. The setup wizard runs several at once, and all runs happen inside one request, so many prompts on many engines can take several minutes. Running in smaller batches is easier to follow.
What it costs
Check yourself
0 of 3 answered.
Progress is saved in this browser.
Put the lesson to work
Paste your site and get a free SEO, AEO and GEO audit with the fixes, then come back to the next lesson.