Category Intelligence

The Guide

How AI systems decide who gets recommended.

This is the framework behind every report we run, laid out in full, free. You can do the self-check yourself. Most founders do, find a gap, and then want the deeper version, which is what the report is for.

The Framework

01

Legibility

AI systems repeat what is easy to summarize confidently: a clear one-line description of who you serve and what you replace. Vague or clever positioning gets paraphrased inconsistently, or dropped.

02

Category default

Once a system maps you to a category, that description repeats across related questions. If a competitor claimed the clear framing first, you inherit the harder position by default.

03

Reference density

Recommendations lean on recognizable references, third-party mentions, comparisons, and reviews, not just your own site copy. Thin third-party presence reads as low confidence.

04

Source attribution

Each system draws on a different mix of sources. A system that leans on documentation and editorial coverage will describe you differently than one that leans on review platforms.

05

Consistency across systems

A one-off strong answer from a single system is not a signal. What holds across ChatGPT, Claude, Gemini, and Perplexity, repeated over multiple phrasings, is.

Run This Yourself

A five-step self-check, before you pay anyone.

01

Ask the question you think buyers ask

Pick the 3-4 phrasings a real buyer would type, not your category jargon, and run them across ChatGPT, Claude, Gemini, and Perplexity. Note who shows up and in what order.

02

Check whether you're described correctly

When you do appear, is the description accurate? A wrong or dated description is often worse than absence, since it means the system formed an opinion from stale or incomplete sources.

03

Look at what the systems cite

Where models will show sources or you can infer them from phrasing, note whether they're your own pages, reviews, comparisons, or press. That tells you what to build more of.

04

Compare against the competitor that beats you

Run the identical query and note the specific language used to describe the competitor. The gap between their description and yours is usually the actionable part.

05

Re-run monthly, not once

Models update. A one-time check tells you where you stand today; the pattern across months tells you whether what you changed actually moved anything.

Common Mistakes

Optimizing for one system

Fixing your ChatGPT answer while ignoring Claude or Perplexity address a fraction of where buyers now research. Each draws on a different signal mix.

Treating one query as the answer

A single prompt is one moment, one phrasing. What holds across many phrasings and repeated runs is the actual signal; anything else is noise.

Rewriting copy without adding references

Positioning language on your own site is the weakest signal available. Third-party mentions, comparisons, and reviews carry more weight than your homepage does.

Waiting for a deal to notice

Most founders find out they're being left off the list when a deal mentions a competitor by name. By then, the pattern has usually been running for months.

Where This Ends

The self-check tells you if there's a gap.
The report tells you why.

Everything above you can run by hand in an afternoon. What it won't give you is the pattern: 20 to 30 queries across four systems, the specific competitor signals driving their recommendations, and a prioritized list of what to change first.

That's the difference between knowing you have a problem and knowing exactly what to fix. The report is $199, one-time, delivered in 48 hours.