The generative AI search performance audit: 20 data points to check before you invest in a tracking platform

Key takeaways (summary) {#key-takeaways-summary} - Run a manual audit before buying a tracking platform. Google Search Console now has free generative AI performance reports for all websites, and your own server logs tell you more about your AI readiness than most paid dashboards. - Citation behavior changes overnight. ChatGPT's average citations per response dropped roughly 27% around the GPT-5.3 rollout in March 2026, and Reddit's citation share collapsed 86% in a matter of days in August 2026. A single-snapshot audit will mislead you. - You probably need fewer prompts than you think. Practitioner benchmarks suggest 50 to 150 prompts per topic is plenty for statistically meaningful data at the topic level. - Product pages, not blog posts, are winning citations. Product pages made up roughly a third of ChatGPT citations in July 2026, nearly double their March share. Check whether your format mix matches what engines actually cite. - Ads now appear in about 30% of ad-bearing windows of ChatGPT search responses. If competitors can buy placement on your prompts, that changes what a tracking platform needs to monitor. ## Why audit before you buy {#why-audit-before-you-buy} The AI visibility tool market moved fast in 2026, and the pricing reflects that urgency. Ahrefs charges $199 per month per platform index, on top of a base plan. Semrush dropped its free tier after the Adobe acquisition closed. Enterprise players like Brandlight run up to $25,000 per month. Before you commit to any of that, you should know what you actually need to measure. I've seen too many teams buy a platform first and discover their problems second. The usual discoveries: their robots.txt blocks half the AI crawlers, their product pages are invisible while their blog gets cited for everything, or their