AI search is rewriting your traffic graph in real time — here's what we do for citability

Two years ago, the question was 'how do we rank.' Today, the question is 'how do we get cited.' AI Overviews now appear on roughly a third of commercial Google queries. ChatGPT, Perplexity, Claude, and Bing Copilot are quietly becoming top-of-funnel discovery surfaces — not for clicks (yet), but for the framing buyers walk into your site already holding. If you're not citable at passage level, you're not in the conversation that decides what they buy.

The shift: from ranking to citation

Classical SEO competes for ranked positions in a list of links. Generative Engine Optimisation (GEO) competes for citations inside AI-generated answers. The two share technical foundations — crawlability, indexability, schema, Core Web Vitals (now including INP, which replaced FID in 2024) — but diverge sharply on output. Ranked links measure clicks. Citations measure presence in the answer the buyer reads before deciding whether to click anything at all.

Brands that win at GEO in the next 18 months will compound an unfair advantage that's hard to claw back. Brands that ignore it will lose discoverability in surfaces buyers are increasingly using to research before they ever hit Google. The leading-indicator window is open now and won't be open in 2027.

What 'citable' actually means

AI engines cite content for two reasons: it answers the question precisely, and it's structured in a way the model can chunk cleanly. The second is where most content fails. A 2,000-word essay that buries the answer in paragraph 7 will not be cited even if the answer is correct. A 200-word section with a clear definition, a numbered list, and an explicit attribution will be cited 5× more.

The five citability factors

1. 01Definitional clarity — does the page open with a 1-paragraph plain-language definition of the topic? (AI engines cite definitional content ~3× more than narrative content.)
2. 02Passage chunkability — are the answers to specific questions findable in 100–300 word self-contained passages? (Q&A formatting + descriptive H2/H3 hierarchy compounds this.)
3. 03Entity richness — does the content name specific tools, frameworks, products, methodologies, dates? (Entities are how the model verifies the page is authoritative on the topic.)
4. 04Schema completeness — FAQPage, HowTo, Article, Organization, sameAs, speakable. (AI engines lean heavily on structured data when deciding what to cite.)
5. 05Author signal — is there a named human author with a Person schema and a credible bio? (Anonymous content is increasingly suppressed; bylined content increasingly favoured.)

The five-step GEO audit we run on every engagement

1. AI-crawler access map

Many sites either block AI crawlers in robots.txt by default (losing discoverability) or allow all of them with no policy (losing control). The right answer is per-crawler: GPTBot, ClaudeBot, PerplexityBot, Google-Extended, Bingbot all have distinct identities and different value to your brand. Audit the existing robots.txt + per-bot HTTP headers and decide explicitly.

2. llms.txt + AI-crawler policy

llms.txt is the emerging standard for telling AI systems what your site is about and which surfaces matter. It's not a magic SEO trick — it's a structured summary in the root of your domain that helps AI engines build accurate brand representations. Brands without llms.txt are increasingly outliers in 2026.

3. Passage-level citability scoring

Score every page against the 5 factors above. Most sites land in the 30–55% citability range — meaning every page is doing 30–55% of the work it could be doing. Lift to 70%+ tends to compound: better passages → more citations → more authority signal → more citations.

4. Schema rebuild for AI discoverability

Article schema alone is 2018-tier. Modern programs deploy FAQPage on Q&A content, HowTo on procedural content, Product + Offer on commercial pages, Organization + sameAs for brand-authority signal, and Person + Author for byline signal. Speakable schema is increasingly relevant for voice surfaces.

5. Editorial calendar tied to AI-citation tracking

Once the foundation is in place, the calendar becomes the engine. Brand-named queries should be cited within 14 days. Category-defining content should be cited within 30. Track citations per-platform and adjust content production accordingly.

What to expect: the lag is short

Unlike classical SEO (where lifts compound over 6–12 months), AI-citation lift is fast. A page rewritten for citability is typically cited in ChatGPT and Perplexity within 30 days of publishing. AI Overviews placement takes 60–90 days. Across an engagement, we typically see brand-named query coverage move from <20% to 60–80% within a quarter.

Why most SEO teams haven't started

Three reasons. First, it doesn't fit the existing reporting (rankings, traffic, conversions) — citations are a leading indicator, not a backward-looking metric. Second, the tooling is fragmented; there is no canonical 'AI citation tracker' the way there is for SERP. Third, the work is mostly editorial and structural — not the link-building, keyword-density, schema-tweak loop most SEO teams are organised around.

Want a citability scorecard for your site?

We run AI-citability scoring + llms.txt + schema audit as a one-week sprint inside the SEO & GEO audit. Output: every page scored, prioritised, with the rewrite plan. $5K, refundable, yours regardless.