Field Notes · June 2026

AI search visibility: getting cited, not just ranked.

A practical playbook for showing up inside ChatGPT, Perplexity, Gemini, and Google AI Overviews — written from the studio bench, not the SEO blogosphere.

The search box is splitting in two. One half is still ten blue links; the other half is a paragraph an AI wrote for you, with three or four tiny citations underneath. If your brand isn't inside that paragraph, you're invisible to a growing share of buyers — and traditional SEO tactics, on their own, won't put you there.

This is the workflow I use with clients to earn citations in AI answers without abandoning the SEO fundamentals that still drive ranked traffic.

1. Decide what you want to be cited for

AI engines reward specificity. "Marketing agency" is a paragraph nobody writes; "AI-powered creative production for FMCG brands in Pakistan" is a sentence a model can lift verbatim. Pick three to five claim territories — the categories, geographies, and outcomes you want to own — and write them down before you touch a single page.

2. Let the AI crawlers in

Most major AI bots respect robots.txt. If you haven't explicitly allowed them, you may be silently excluded from the index they answer from. The bots to know in 2026:

  • GPTBot — OpenAI / ChatGPT Search
  • OAI-SearchBot — ChatGPT live retrieval
  • PerplexityBot — Perplexity index
  • Perplexity-User — Perplexity live fetch
  • Google-Extended — Gemini & AI Overviews training
  • ClaudeBot — Anthropic
  • Applebot-Extended — Apple Intelligence

Allow each one in robots.txt, and publish an llms.txt at the root with a short site summary plus a link to every page worth citing. Treat it like a press kit for machines.

3. Write in claim-first paragraphs

LLMs pull short, self-contained sentences. Lead every section with the answer, then support it. Anti-patterns to delete:

  • Throat-clearing intros ("In today's fast-paced world…")
  • Buried numbers ("we grew it significantly")
  • Floating pronouns with no subject in the same sentence

Replace with: a one-sentence claim, a specific number, the year, the source. That's the shape models cite.

4. Ship structured data on every page

JSON-LD is how you tell AI engines what a page actually is. The four schemas that move the needle:

  • Organization / ProfessionalService sitewide
  • Article on every post, with author and datePublished
  • FAQPage on any page answering discrete questions
  • BreadcrumbList on deep routes

5. Build entity authority, not just backlinks

AI engines resolve people and companies as entities and weight citations from sources that already know who you are. Practical moves:

  • Claim and complete your Wikidata, Crunchbase, and LinkedIn entries
  • Use the same name, bio, and headshot across every profile
  • Link sameAs in your Organization JSON-LD to every profile
  • Guest-author or get quoted on sites that already get cited

6. Answer the questions, literally

Pull the long-tail questions from Perplexity, AlsoAsked, and your own sales calls. Create one page per high-intent question, with the question as the H1 and a 40–60 word direct answer in the first paragraph. This is what gets lifted into AI Overviews.

7. Measure citations, not just clicks

AI traffic shows up in analytics as referrals from chat.openai.com, perplexity.ai, gemini.google.com, and increasingly direct/organic with long, conversational queries. Weekly: ask the same five buyer questions in each engine and log which brands get cited. That diary is your real ranking report in 2026.

The short version

Decide the claims you want to own. Let the bots in. Write so a model can lift you in one sentence. Mark up the page so it knows what it just read. Then build the entity around it. The brands earning AI citations aren't writing more content — they're writing quotable content, and shipping the plumbing underneath.

Want help wiring this into your site? Start a project →