How to choose the right prompts for LLM tracking

In classic SEO, keyword volume pointed at a plan. In AI search, one buying decision spans many prompts — and personalization means the same words can yield different answers for different people. Volume alone won't cut it.

Three coverage axes

• Journey stage — awareness, consideration, evaluation/purchase
• Customer context — segments, constraints, use cases
• Geography — if you sell in multiple markets

Awareness

Catch category fears before brand shortlists form. Pull objections from sales calls, Reddit, and 'what concerns do people have about [category]?' Turn them into prompts: reliability, cost, fit for families, cold weather, and so on.

Consideration

Track broad 'best X' prompts and segment variants ('best X for small teams'). Include plain category nouns — people still type them.

Purchase and evaluation

Ecommerce needs 'where to buy' coverage. Everyone needs brand evaluation tracked separately ('Brand A vs Brand B') so 100% visibility on branded prompts doesn't inflate your averages.

Fast start without deep discovery

1. Convert top non-branded GSC queries into natural questions
2. Read the site nav and pricing tiers for segments
3. Draft 10–20 awareness + 20–30 consideration prompts
4. Track 30 days in Anny, then fill gaps where competitors dominate

Winning teams cover the journey strategically. They don't track everything — they track the moments that decide deals.