What does Google itself say about SEO in the AI era?
Google's May 2026 guidance on optimizing for generative AI features is unambiguous: there is no separate AI optimization discipline, no special markup, no secret file that gets you into AI Overviews. Pages must be indexed, crawlable, and rankable in regular Search first — AI features select from the same corpus. In Google's words, AEO and GEO are 'still SEO'.
That cuts both ways. It means your existing SEO foundation is not wasted — it is the qualifying round for AI visibility. But it also means sites that never did SEO well are invisible twice over: absent from shrinking classic results and absent from the AI answers replacing them. The fundamentals did not die; they became the entry fee.
What actually changed about search behavior?
Zero-click is the headline shift: AI Overviews and chat interfaces answer informational queries in place, so top-of-funnel content earns dramatically fewer visits even at top rankings. Meanwhile a growing share of high-intent research — 'best X for Y', vendor comparisons, pricing questions — happens inside ChatGPT, Perplexity, and Gemini, where the outcome is a mention or citation, not a ranked link.
What did not change: commercial and transactional queries still convert through classic results, local search still drives foot traffic, and referral traffic from AI surfaces — while smaller in volume — converts at notably higher rates, because a user arriving from an AI recommendation has already been pre-sold by the answer that named you.
What SEO tactics still work — and which died?
Still working, arguably harder than ever: original data and first-hand experience (the March 2026 core update rewarded information gain — content adding genuinely new knowledge), named authors with verifiable credentials, technical health and Core Web Vitals, structured data (roughly 65% of AI-cited pages carry schema), and digital PR — third-party corroboration now feeds both link equity and LLM trust.
Dead or dying: mass-produced synthesis content restating what already ranks (AI engines synthesize better than your content mill), anonymous-byline blogs, exact-match keyword stuffing, and traffic-only reporting that ignores where answers actually happen. The pattern across every casualty is the same: anything an LLM can generate itself no longer earns a citation from one.
- Works: original data, named expert authors, schema, digital PR, technical health
- Works: answer-first formatting — direct 40-60 word answers under question headings
- Dead: synthesis content, anonymous bylines, keyword stuffing
- Dead: measuring success in rankings and traffic alone
How should you measure SEO success now?
Keep the classic dashboard — rankings, organic traffic, conversions — and add an AI-visibility layer: mention rate and citation rate across a fixed set of prompts your buyers actually ask in ChatGPT, Perplexity, Gemini, and AI Overviews, tracked monthly. Add referral traffic from AI surfaces as its own segment, and weight it by its higher conversion rate rather than dismissing its volume.
The strategic reframe: search visibility was always a proxy for 'do buyers encounter us while deciding'. In 2026 that encounter happens across two surfaces — ranked results and synthesized answers — and measuring only the first is how companies convince themselves SEO 'stopped working' while their competitors quietly become the default answer in the second.
What should you do differently — and where does Chalk Labs fit?
The 2026 playbook: keep technical SEO and indexability pristine (the entry fee), restructure key pages answer-first with question-form headings, publish original numbers worth citing, put named experts behind everything, build third-party presence — listicles, reviews, PR — because LLMs recommend brands they see corroborated across sources, and track AI mentions alongside rankings.
This integrated practice — SEO, AEO, and GEO as one discipline — is exactly what Chalk Labs sells, with retainers from around $3,000 per month. We eat our own cooking: this page is built answer-first, schema-marked, and expert-bylined precisely because that is what earns both the ranking and the citation. If you found it through either surface, the method just demonstrated itself.
Questions we hear about this
No — it is the qualifying round for AI visibility. Pages must rank to be cited. What died is traffic-only SEO: informational clicks shrank, so strategy must target citations and mentions alongside rankings.
They are sequential, not alternatives. Indexability, technical health, and rankable content come first because AI engines select from that corpus. GEO tactics — entity work, citations, answer formatting — build on the SEO base.
Only content with information gain: original data, first-hand results, expert judgment. Synthesis posts restating what ranks are exactly what AI generates for free. One citable original page beats twenty rewrites.
Similar to before: 3-6 months for meaningful movement, faster in low-competition niches. AI-citation visibility often moves within 6-12 weeks once foundations are fixed, since answer engines refresh on crawl cycles.