Free AI prompt research tool
Which AI prompts should you track?
Enter your domain. We read your site, check search demand in your market and write the questions buyers ask AI assistants while they choose. You get a scored prompt set sized to your budget, with the reason and the source behind every prompt.
Free, no card, no account, about two minutes. Prompts are written for:
- ChatGPT
- Perplexity
- Gemini
- Google AI Mode
- Google AI Overviews
What you get
A prompt set you can defend in a client meeting.
Not a list of plausible questions: every prompt carries the buyer it serves, the stage it sits at, the demand behind it and the reason it was kept.
Buyer prompts, one per intent
Conversational prompts for ChatGPT, Perplexity and Gemini, plus a short search form for AI Overviews. Rewordings of the same question are merged.
Search demand behind each prompt
The nearest real keyword and its monthly volume in your market, so you can see which questions have proven interest.
A balanced mix by prompt type
Own-branded, competitor-named, shortlist, constraint, how-to and approach prompts, each held to a target share before anything is ranked. The targets follow whether AI answers in your category name vendors or cite sources.
Sized to your answer budget
Pick your plan and engines. The set is redrawn to the number of prompts you can run daily, with the margin of error per engine.
Coverage checks
Every top competitor, priority persona and core product line gets enough prompts to report on, or the report says which is thin.
Reasons, sources and CSV
See why each prompt made the cut and which sources back it. Filter by topic, persona or stage, generate more, correct the brand model and redraw, and export the set into any prompt tracker.
How it works
The four steps a good analyst would take.
AEO is about winning the direct answer. GEO is about being cited or recommended inside AI answers. Both depend on tracking the questions buyers actually ask, so the tool follows the same method.
Read your site
Home, product, pricing, comparison and integration pages give your personas, product lines, competitors and what you do not offer, and show whether you have content that AI answers cite.
Measure demand
Search volume and Google’s People Also Ask questions in your market show which topics people already ask about.
Write and filter
One conversational prompt per buyer intent. Off-offer, leading and “top 10” style prompts are removed, with the reason kept.
Score and fill the mix
Each type gets its share first, then prompts are ranked by value, awareness stage, gap, leverage, fit and evidence.
Sample result
What a prompt set looks like: a consumer VPN service, United States.
100 candidates written from 20 pages of the brand’s site and United States search demand. 16 merged as rewordings, 84 scored, 38 fit a 6,500-answer monthly plan on 5 engines.
- 100
- candidate prompts written
- 16
- merged as rewordings of the same question
- 84
- distinct buyer intents scored
- 38
- in the core set
- 68%
- unbranded (target 60 to 75%)
- ±9.2
- points monthly margin per engine
- Score 0.80
Is a paid VPN worth it over a free one for privacy, and which paid ones are good?
Approach comparisonPrivacy-conscious usersbest vpn network free · 33,100/mo - Score 0.88
Is Competitor A really $2.99 a month, or are there cheaper VPNs that are just as good?
Competitor-namedMost aware · choosingcompetitor a pricing · 5,400/mo - Score 0.79
Which VPN app for Android is safest on public Wi-Fi without draining my battery?
Select (constraint)Privacy-conscious usersbest vpn software android · 3,600/mo - Score 0.77
Which VPN app is best for my Android phone when I use public hotspots?
ShortlistPrivacy-conscious usersbest vpn software android · 3,600/mo - Score 0.69
Will Netflix ban my account if I use a VPN to watch other libraries, and how do I avoid errors?
Problem and how-toProduct aware · checkingnetflix vpn service · 590/mo - Score 0.55
Is [your brand] any good for streaming Netflix and Hulu, and what do users say about it?
Own-brandedReported apartnetflix vpn service · 590/mo
Each prompt also gets an AI Overviews search form, a persona, an awareness stage, a topic and a full score breakdown. Brand and competitor names in the sample are replaced with generic labels.
Built differently
Most free tools write a list of questions. This one decides which are worth paying to track.
A branded prompt names you almost every time, so too many of them inflate your visibility score. Shares are fixed before ranking so no type can dominate, and the set is sized to the answers you can afford.
The six prompt types in a balanced set
- Own-branded · 10–15%“How much does [your brand] cost?” Reported apart from the headline.
- Competitor-named · 15–25%“What are good alternatives to [competitor] for streaming?”
- Shortlist · 25–30%“Which VPN services have a verified no-logs policy?”
- Select (constraint) · 15–20%“Which VPN fixes strict NAT on a games console without lag?”
- Problem and how-to · 10–15%“How do I torrent without ISP warning letters?”
- Approach comparison · 0–10%“Is the VPN in my antivirus good enough?”
How many prompts should you track?
Enough to read a real change. A monthly number needs about 30 unbranded prompts per engine; below that the margin of error grows past what most changes can clear. One answer is one credit, and each prompt runs once a day on each engine.
| Plan (answers a month) | 1 engine | 2 engines | 3 engines | 4 engines | 5 engines |
|---|---|---|---|---|---|
| Standard · 6,500 | 190 | 95 | 63 | 47 | 38 |
| Growth · 18,000 | 500 | 250 | 166 | 125 | 100 |
Prompt slots per engine after reserving answers for the pilot, rotation and top-ups. At 38 core prompts (33 unbranded) the monthly margin is ±9.2 points per engine; at 100 (87 unbranded) it drops to ±5.7.
Prompt research vs keyword research
Keyword research tells you what people type into Google. Prompt research tells you which questions to put in front of AI engines. You need the first to do the second well.
| Aspect | Keyword research | Prompt research |
|---|---|---|
| Unit | A short query: “vpn for streaming” | A full question with context: “Which VPN works with my streaming service without proxy errors?” |
| Demand data | Published search volume | No engine publishes prompts, so demand is borrowed from the nearest keyword |
| What you measure | Position in a ranked list | Whether answers mention or cite you, over many runs |
| Rewordings | Tracked as separate keywords | Merged: they measure the same thing |
| Set size | Hundreds to thousands | Sized to an answer budget and a margin of error |
In one 330-answer test, a single branded prompt produced 30 of a brand’s 34 mentions. Visibility is an average over the prompts you track, so the prompt set decides the score before any engine answers.
Next steps after you have your prompts
- Prompt tracking runs your set daily across the AI engines, with mentions and citations per prompt.
- ChatGPT visibility tracker checks whether ChatGPT mentions your brand for a prompt, free.
- AI Overviews tracker shows which sources Google’s AI Overviews cite for your topics.
Questions
Prompt research questions
What the tool measures, how it chooses, and what it will not claim.
What is prompt research?
Prompt research is finding the questions buyers ask AI engines about your category, then choosing which of them to track. It starts from keyword research, because no engine publishes what its users type, and turns those topics into full conversational questions.
How do I choose prompts to track for AEO and GEO?
Start from your own site and real search demand, write one prompt per buyer intent, remove leading or off-offer wording, then hold each prompt type to a share before ranking by value and awareness stage. For AEO, favour questions with a direct answer. For GEO, favour shortlist and comparison questions where engines name and cite brands. The tool reads which your category needs: where AI answers name vendors the set follows the buyer journey, and where they explain and cite sources it follows the questions your content answers.
How many prompts should I track?
Enough to read a real change: about 30 unbranded prompts per engine for a monthly number. A small budget is better spent on more prompts on one engine than a few prompts on five.
What is prompt volume, and can I trust it?
Prompt volume is an estimate of how often a question is asked in AI engines. Engines do not publish it, so every figure is modelled. This tool shows the nearest real keyword and its search volume instead, which you can check yourself.
Why not just track my top keywords?
Keywords are short and AI prompts carry context like budget, device or team size, which changes which brands are named. Keywords are the starting point; the prompt is what you measure.
Why so few branded prompts?
A prompt that names you returns your brand almost every time, so it inflates your visibility score. Branded prompts are kept to 10 to 15% (less where AI answers cite sources rather than name vendors) and reported apart as an early warning on pricing and facts.
Which AI engines should I track?
Start where your buyers ask: usually ChatGPT, then Google AI Overviews or AI Mode, then Perplexity or Gemini. Add an engine only when each one can still get about 30 unbranded prompts.
What is the difference between prompt research and prompt tracking?
Prompt research picks the questions. Prompt tracking runs them on AI engines every day and records whether you are mentioned or cited. This tool does the first; our prompt tracking does the second.