AI Summary
AI search's non-deterministic nature renders traditional ranking metrics obsolete, with only 30% of brands remaining visible across consecutive AI answers. New metrics like citation rate (SaaS leaders target 20-30%), share of voice, answer set membership, and sentiment are crucial for tracking brand visibility. Tools like the open-source GEO/AEO Tracker can help measure these metrics across multiple AI models.
Only 30% of brands stay visible from one AI answer to the next. Just 20% remain visible across five consecutive runs of the same query. AI search is non-deterministic by design – which means the SEO industry’s old “ranking position” metric is dead. Here are the metrics that replace it.
Why traditional rank tracking breaks for AI search
Three structural reasons:
- AI answers are generated, not retrieved. Re-running the same query produces variation in citation order and inclusion.
- There is no “position 1.” Citations are typically a set of 3-7 unordered references.
- Personalization is heavier. ChatGPT and Perplexity factor user history; Google AI Overviews vary by location and account.
The four metrics that matter
- Citation rate. For a target prompt, run it N times (10-25) per AI engine and record the % of runs that cite your domain. SaaS leaders target 20-30% as the threshold of meaningful AI visibility.
- Share of voice. Of all citations across N runs, what % were yours vs. each named competitor? This is the single best leading indicator of brand share in your category.
- Answer set membership. Beyond citation, is your brand mentioned by name in the answer text (with or without a citation link)? Mentions without links still build brand association.
- Sentiment and positioning. When you are mentioned, are you positioned as the leader, an alternative, or a “not recommended” option? This requires reading the answer text – manual or via LLM-assisted analysis.
Building (or buying) the tracking infrastructure
Three options, ranked by cost and control:
- Open source. The GEO/AEO Tracker is a local-first dashboard tracking citation rate, share of voice, sentiment, and answer set membership across ChatGPT, Perplexity, Gemini, Copilot, Google AI Mode, and Grok. Free to self-host; you bring API keys (~$10/month total cost).
- Mid-market SaaS. Profound, Peec AI, Otterly, AirOps. $99-$500/month. Convenient, less customizable.
- Enterprise. Quattr, Brand Radar (Ahrefs), Conductor. $500-$2,000+/month. Strong for agencies managing many brands.
Whatever you choose, the rule is the same: measure weekly, segment by prompt category, and treat the trendline (not the absolute number) as the primary signal.
Frequently Asked Questions
How many times should I run the same prompt to get a reliable measurement?
10-25 runs per prompt per engine. Below 10, variance dominates. Above 25, marginal precision is small. For high-stakes prompts, run weekly and look at the trendline.
Which AI engine should I prioritize tracking?
Start with the engine your buyers actually use. For B2B research-driven categories, prioritize Perplexity and ChatGPT. For consumer information queries, prioritize Google AI Overviews. For developer audiences, add Copilot.
Can I track AI visibility without paid tools?
Yes – manually for small prompt sets, or via the open-source GEO/AEO Tracker for systematic tracking. Manual works up to ~10 prompts; beyond that, automation becomes essential.
Want this implemented for your brand?
I help growth-stage companies own their category in AI search. Set up AI visibility tracking.