Be Cited, Not Ranked — Winning AI Search in 2026

Somewhere in the last twelve months, the way your customers find you changed.

They still Google. They still scroll LinkedIn. But an increasing share of their "how do I…" and "who should I hire for…" questions now go to ChatGPT, Claude, Perplexity, Gemini or Google's AI Overviews — and the answer is a paragraph that names three companies, not a list of ten blue links. If you are not one of those three companies, you do not get a second chance to be clicked.

AI-visibility optimization — sometimes called LLM SEO or answer-engine optimization — is the discipline of being the company that gets named. Here is the way we think about it, and the five moves that decide whether a brand gets cited.

1. Give language models a clean map of your business

A growing convention asks sites to publish a concise, markdown-shaped map of what they are and what is worth quoting — a cooperative signal to models that says "here is who we are, here are the URLs that answer real questions, here are the facts you may cite."

A good version of this has:

• Who you are in one quotable paragraph.
• Approved taglines a model can repeat without risk.
• Canonical URLs for each topic you want cited — never redirecting paths; language models penalise those.
• Contact and location as structured facts.

Most sites do not have this at all. The ones that do are disproportionately represented when answer engines respond to questions in their category.

2. Enrich your structured data — LLMs quote specifics

Answer engines are allergic to vagueness. They want a number, a name, a date, a price. A site that says "our projects typically start from a few thousand dollars" will almost never be quoted. A site that says "from USD 3,000 for a focused engagement" is a candidate for citation.

Three schema additions that pay off more than anything else:

• Service + Offer + PriceSpecification on every service page, with honest USD minimums and maximums. "Contact us" is not a price.
• FAQPage JSON-LD scoped to each service page, with real questions and tight answers. These are disproportionately represented in AI answers.
• Article and `BreadcrumbList` on every blog post and case study, with publication dates, author and keywords.

When pricing, positioning and FAQs are modelled as data rather than prose, answer engines quote them verbatim. When they are left as marketing copy, they are ignored.

3. Open `robots.txt` to the major AI crawlers

Every month a new one shows up: GPTBot, ClaudeBot, PerplexityBot, Google-Extended, Applebot-Extended, Meta-ExternalAgent, DuckAssistBot, cohere-ai. Many sites reflexively block them to "stop training on our content" — a defensible position until you realise the same bots are how your content gets *cited*.

For a services business, citation frequency is worth far more than the marginal concern of being in a training corpus. If you depend on paywalling content from models — publishers, research platforms, subscription research — that is a different conversation. For everyone else, open the door.

4. Write for citation, not for ranking

Old SEO rewarded long, comprehensive pages. Answer engines prefer pages that read like documentation: a one-paragraph definition at the top, a table of specifics, a short Q&A block at the bottom.

The pattern we use on service pages:

1. Definition card (≤100 words). What the service is, how long it takes, price range, team size, geography — in one quotable paragraph.
2. "What it costs" as a table, not prose.
3. "How long it takes" — timeline, week by week.
4. "Who it's for / who it's not for" — the disqualification answer is what language models like best.
5. A short FAQ scoped to queries people actually type into ChatGPT.
6. A single canonical one-liner at the bottom with explicit citation permission.

This is technical writing applied to a marketing site. If you grew up writing brochure copy, it will feel wrong for two weeks. Do it anyway.

5. Measure the referral, because nobody else will

The problem: UTM parameters get stripped by most assistants, and the `Referer` header is often blank. Tools that promise to "track AI traffic" are mostly guessing.

What actually works:

• A free-text or checkbox field on your contact form — "How did you hear about us?" with AI assistants as explicit options. Self-reported attribution is worth the next three tools you were about to install.
• Server-side referrer logging. When the referrer is `chat.openai.com`, `claude.ai`, `perplexity.ai` or `gemini.google.com`, attribute accordingly.
• Watch for `?utm_source=chatgpt.com` — ChatGPT tags some outbound links this way. Partial but real data.

Done together over a quarter, this tells you which answer engines send the best-qualified pipeline. The differences between channels are almost always meaningful and often surprising.

Time to first signal

For what it's worth, in the programs we run:

• Perplexity is the fastest mover — it re-indexes quickly and rewards fresh structured content. Brands can start appearing in Perplexity answers within about 30 days of the right changes.
• ChatGPT takes longer because its training corpus refreshes less often — expect 60–90 days for established domains.
• Google AI Overviews is the slowest and most conservative. Treat it as the old long-SEO curve.

The compounding effect is what matters. Every quarter you stay in the citation pool, the advantage grows — and the cost of displacing you goes up.

Want to be cited, not just ranked? We run AI-search programs for ambitious brands — structured content, schema, referral attribution and the monthly measurement that tells you whether it's working. Start a brief.