5,507,885unique real prompts
6,900,514privacy-cleared rows
1,009,483commercial-intent prompts
3industry reports published
Main findings
What people use AI for
Most prompting is not commercial. Sorting that out first is what stops a
share-of-voice number being quoted against the wrong denominator.
Commercial intent is 15.6%
of cleared prompts. That is the pool worth optimising for.
Short questions and long consultations are different behaviours
Averaging these together is how you get a meaningless "average prompt length". They are
two distinct modes and they need different content.
What this is not
These are prompt-phrasing counts from a fixed corpus. They are not search volume,
not market size, not demand forecasts and not AI answer share of voice. Percentages are
shares of this corpus. Small counts are labelled directional and should be verified
before they drive spend.
Leaderboard
The commercial prompt types, ranked
Within commercial prompting, this is what people are actually asking. Read it as
the jobs AI is being hired to do when someone is shopping.
Buying stage
A prompt can carry more than one stage, so these rows sum to more than the commercial total. Stage is a relabelling of the commercial prompt type above, not an independent measurement: see the Method tab.
What they constrain the answer by
The constraint someone attaches to a prompt is the thing your page has to answer.
Feature and price dominate, and location matters far more than most content plans assume.
GEO tracking prompts
How competitor comparison is actually phrased
The real shape of comparison prompts, for building a competitor-tracking portfolio
grounded in observed phrasing rather than invented examples.
Real opening phrases, by commercial type
Candidates pass the same rejection filter the industry report uses: no PII, no content-generation, no supply-side, must carry a research cue. Automated bot loops and
role-label parsing artifacts (a corpus-wide scan found 289,543 near-duplicate prompts)
are excluded before these templates are built.
Which brand should I pick
25,935 matched prompts
what is the best way 1,373what s the best way 535what are some of the 196what would be the best 168what are the top 10 167what are the top 5 107
Tell me about this category
16,415 matched prompts
within this same context what 79what is the difference between 60what are some of the 40what is the best way 32what do you know about 26what would be a good 22
Find me someone local
15,910 matched prompts
in general is a an 94what are some of the 60who are all of the 39what is the weather in 37what is the difference between 37what is the weather like 37
What does it cost
9,272 matched prompts
how much does it cost 105what is the price of 64what is the cost of 54how much would it cost 49generate a short aesthetic cute 31what is the most expensive 28
Compare these two
6,744 matched prompts
how do you compare to 58compare and contrast the performance 38what is the difference between 36why are you better than 19what are the pros and 17how do you compare with 17
I am ready to buy
6,239 matched prompts
what is the best book 24what is the best way 21do you know the book 20what is the order of 18what is the difference between 13what are some of the 11
What else is there
3,472 matched prompts
what are some alternatives to 37what are the alternatives to 17what are the benefits of 15are there any alternatives to 14is there a way to 14is there an alternative to 13
Is it any good
2,439 matched prompts
what is the difference between 9can you review the class 820 recent empirical review on 6literature review on organizational change 6review and list positives of 6what is rolloff high frequency 6
Can I trust them
1,850 matched prompts
is it safe to use 21is it worth it to 18what is the most reliable 14is it safe to eat 13choose a company and explain 10is it safe to drink 8
By assistant
The only source that knows which assistant a prompt went to
ShareChat is shared-conversation links, split by platform. Read every number here as a
rate, not a raw count: assistant volumes differ by orders of magnitude, so
comparing raw counts would just be comparing ChatGPT's larger user base, not their behaviour.
Comparator phrasing rate, by assistant
Share of each assistant's commercial prompts using this phrasing.
Qualifier mix, by assistant
Industry reports
Published reports
Each report is the same analysis narrowed to one industry, with real example prompts
and a prioritised build list.
Connect Claude
Run this yourself, in Claude
The corpus is hosted privately on Cloudflare R2. The skill pulls it once then runs
locally, so every run after the first is fast and costs nothing.
Three steps
- Install the skill folder, and rclone if you do not have it.
- Drop the read-only R2 credentials at
~/.r2-corpus.conf. Ask Joe for them.
- Run
python scripts/rocket_geo_intelligence.py run --industry <name>.
The corpus downloads itself on first use.
Adding a new industry is one JSON config in configs/. Copy
_industry_template.json, set the terms, run it, publish.
Method and limits
Corpus lanes are kept separate on purpose. Public LLM-chat data is not confirmed
ChatGPT traffic, and traditional-query data is not AI-chat behaviour. "Buying stage"
on the Leaderboards tab is a fixed lookup off the commercial prompt type (verified
100% identical across all commercial rows), not an independent measurement: treat
it as a relabelling, not new evidence. Automated bot loops and role-label parsing
artifacts are excluded from every count on this site; the raw corpus is larger than
what is shown here because of that filtering, which is deliberate.
Corpus: 5,507,885 unique prompts, 6,900,514
privacy-cleared rows.
