What Is a Prompt Set?

In one sentence

A prompt set is a fixed list of questions used to measure AI visibility consistently — the benchmark everything else is scored against.

What Prompt Set means

It should reflect how your category is actually asked about conversationally, which differs substantially from how people type into a search box.

Crucially it stays fixed. Changing questions between reports makes month-over-month comparison meaningless, and it's a common way agencies flatter results.

Why it matters for AI search

Without one, AI SEO is unfalsifiable. Screenshots prove nothing when models return different answers to the same question twice.

A well-built prompt set is also the single most useful artefact from an audit — it's yours to re-run independently at any time.

See it in action

Where the questions actually come from.

Sourcing a prompt set

Benchmark design
Pillar

40–60 category prompts

Search query data

What people already type, rephrased conversationally

Sales conversations

The questions your team fields on every call

Competitor positioning

How rivals frame the category

PAA and fan-out

Sub-questions the topic decomposes into

Comparison intent

X vs Y, alternatives to X

Commercial intent

Pricing, implementation, who to hire

Six sources, one fixed set, approved before tracking starts.

Client approval matters. A set the agency chose alone can be quietly optimized for, which is exactly what makes the number meaningless.

How to get it right

Building a prompt set that measures something

  • Draw from real query data, sales conversations and competitor positioning
  • Phrase questions conversationally, in full sentences, as people actually ask them
  • Get client approval before tracking begins, then freeze the set
  • Run each prompt multiple times per cycle so variance is captured
  • Revisit quarterly rather than ad hoc, and document any change

Common questions

How many prompts should be in a set?

Typically 40 to 60 for a focused category — enough to cover real question clusters, few enough to run reliably at multiple repetitions.

Why does the set have to stay fixed?

Because a changing set makes month-over-month comparison meaningless. It's also the easiest way to make results look better than they are.

Can we build one ourselves?

Yes. Start with the questions your sales team hears most, phrase them conversationally, and run them consistently. The discipline matters more than the tooling.

These come up alongside Prompt Set constantly.

Free 12-point check

Send Ron your site

Two fields. We’ll crawl your site as GPTBot, ClaudeBot and PerplexityBot, benchmark a sample of your category’s prompts, and send you what we find.

  • What each AI crawler actually receives from your site
  • Whether models resolve your brand as a real entity
  • A sample of category prompts and who gets cited
  • The three fixes we’d prioritise first

No commitment, no sales sequence. If we’re not a fit, we’ll say so.

We’ll pull your favicon so you know we found the right site.

We reply from a real address. No sequence, no newsletter.