Inside Claude in Delhi: recap of the SaaS SEO Alliance 2026 workshop

Ranking helps you get discovered. Citability helps you get chosen. A brand can be first on Google and still be absent from an AI answer: the answer depends on the question, the evidence retrieved and the claims that sources support.

• Nameable: one consistent brand name, a clear description, stable identifiers.
• Corroborated: independent sources that confirm the same specific fact.
• Extractable: each key claim in a self-contained sentence, backed by a fact or a number.
• Reachable: important pages open to crawlers, current and readable.
• Repeated: the same verified facts, stated consistently over time.

The Google leak describes more than 14,000 attributes, organised here into 38 working dimensions. On the public HouseFresh case, where traffic fell 95% before recovering to 9% of its baseline, the EQOS Forecast multiplicative model returns 0.09, the observed level. Mapping Google and Anthropic signals gives a 63% overlap in our mapping. Every claim is labelled fact, signal or calculation, to show where the evidence ends and the interpretation begins.

The takeaway for your work: many familiar SEO principles still matter (clear entities, accessible pages, specific claims, corroboration), and the extra work happens at answer level.

1. Write five recommendation questions a real buyer would ask.
2. Run the same five questions through two, three or four answer engines.
3. Record the recommended brands, the citations and the source domains.
4. Calculate one citation score per engine: answers that mention or cite your brand, divided by five.
5. Pick the weakest lever, fix it, then repeat the same test.

You leave with a baseline you can measure again after every change.

Two questions came from the room: which LLM brings the best conversion rate, and whether ChatGPT and Claude extract website content the same way. The detailed answers are on the Q&A page.

Read the QA