Anti-Prime Queries

Definition

An anti-prime query is a prompt designed to NOT mention the brand, product, or specific feature you want to measure visibility for. Instead, it describes the buyer's job-to-be-done (JTBD) in their own language, then watches whether the model surfaces the target organically.

The term comes from the Brand Radar dispatch methodology. Most brand-monitoring tools "prime" the model — they ask "What do you know about Notion?" or "Compare Notion and Confluence." The model has already been told the answer is somewhere in the response set, so it almost always returns something. Anti-prime queries strip the priming away: "best knowledge base for a solo dev using AI coding agents." Now the model has to choose what to return without being told what to look for.

Why it matters

The asymmetry between primed and anti-prime surface rates is the single most important signal in AI search measurement.

In the Notion audit (2026-05-23), 5 of 5 platforms described Notion's MCP integration accurately when the query named MCP explicitly. The anti-prime version — 20 cells where queries described what MCP solves without using the protocol name — surfaced Notion on only 3 of 20 cells. 15% organic versus 100% primed.

That gap is the actual buyer experience. Real buyers don't search for "the Notion MCP server" — they search for "how to give Claude context on my docs." If your monitoring tool only runs primed queries, you'll think your brand is well-covered when in fact you're invisible to the buyer language that actually matters.

How to construct anti-prime queries

Three rules:

1. Describe the job, not the tool. Replace "best CRM for SaaS" with "how to track our sales pipeline if we mostly close on Slack." Replace "best AI SEO tool" with "how to find out if ChatGPT recommends my brand."
2. Use the buyer's vocabulary, not yours. Strip out internal jargon and category labels. If buyers say "sticky note app" and you call it "knowledge management," anti-prime the buyer wording.
3. One job per query. Compound questions ("what's the best CRM that's also a project tracker") make scoring ambiguous. Pick one job per cell.

Where Citare uses anti-prime queries

Every Brand Radar dispatch includes both primed and anti-prime cells. The default 50-cell weekly run for a brand uses 10 primed queries × 5 personas + a parallel 10 anti-prime queries × 5 personas. The two ratios get compared in the dispatch report — the wider the gap, the bigger the marketing-distribution problem (regardless of how strong the underlying product is).

Common pitfalls

• Treating anti-prime as the only valid measurement. Both matter. Primed queries measure how well-described your brand is to a model that already knows you exist. Anti-prime measures whether you'll be discovered at all. You want strong scores on both.
• Making anti-prime queries too vague to be useful. "What should I use?" is too vague. "What should a 5-person SaaS team use to give Claude context on their codebase?" is anti-prime + specific.
• Forgetting to refresh anti-prime queries quarterly. Buyer language drifts. The queries that worked 6 months ago may now contain category labels that have themselves become primes.