How to be cited by AI

What does “cited by AI” actually mean?

When a reader asks ChatGPT, Claude, Perplexity, Gemini, or Bing Copilot a question, those engines crawl the open web for sources and surface specific articles as the answer. Being cited means your piece is what the AI quotes — and the reader gets a link back to your work, on your verified domain.

PUBlish tracks two signals: crawls (an AI engine read your piece) and referrals (a real human clicked through from an AI answer). Both surface on /desk/visibility with a per-engine breakdown.

What PUBlish ships for you automatically

Every piece you publish carries the technical infrastructure AI engines need to find, parse, and cite you. You do not have to think about any of this:

• Schema. BlogPosting + Person + Organization + BreadcrumbList + Speakable JSON-LD on every piece. Your byline becomes a Person entity with knowsAbout, sameAs, and worksFor fields the engines cluster on.
• Canonical credit. When you verify your business website, the canonical URL on every piece points home to your domain. Google credits your site, not PUBlish. You keep the rank you earn.
• llms.txt manifest. A live AI-crawler-specific feed at /llms.txt with the platform overview, your author entry, and your latest pieces. AI engines ingest it on every crawl.
• Permissive robots. GPTBot, ClaudeBot, PerplexityBot, Google-Extended, and CCBot are explicitly allowed. Nothing about your work is hidden from the engines that matter.
• Sitemap + image sitemap. Every published piece + author profile + journal entry indexed with images, priority signals, and last-modified timestamps.

The seven moves you control

The schema is necessary but not sufficient. What earns the citation is the shape of what you wrote. Seven moves, in order:

1. 1Write to answer one question. AI engines pick answers, not opinions. The strongest signal is a piece that answers a specific question a reader is actively typing. Write the title as the question your ideal reader Googles at 11pm.
2. 2Lead with an H2 that IS the question. Open the body with an H2 phrased as the exact question. Then answer it. The H2 becomes the citation anchor — AI engines match queries to headings before they match prose.
3. 3Answer in the first 60–100 words. The first paragraph is the citation window. Put your one-sentence answer first. The rest of the piece is depth, evidence, examples. If the answer is buried under throat-clearing, a different writer's top-of-piece answer wins the citation.
4. 4Back every claim with a dated stat or named source. "Conversion rates dropped 14% in Q1 2026" cites. "Conversion is falling" does not. Numbers and dates are the signals AI engines weight as factual. If you can name the source (your own client work, a public study, a verifiable event), the citation likelihood compounds.
5. 5Break depth into FAQ-style H2s. After your lead answer, split the rest into "## How long does X take?", "## What about Y?", "## Why does Z happen?". Each becomes a separately-citable unit. The same content as flat prose gives the engine one unit; as Q&A it gives three.
6. 6Be specific to your niche and your location. "How design studios in Manchester bill for revisions" beats "How designers bill for revisions." Specificity is topical authority. Topical authority is what AI engines reward. Pick a topic you can own and ship five pieces on it before moving on.
7. 7Verify your identity. A green identity tick on your byline tells AI engines this is a real, named source. The schema gains a hasCredential field. Add one verified social profile (rel="me" cross-check) plus your business domain, and the citation surface compounds.

What NOT to do

• Do not write generic tips lists. “Five marketing tips” loses to “How a London PR firm cut its client onboarding from three weeks to four days”.
• Do not bury your answer. The first 100 words is the citation window — not the conclusion paragraph.
• Do not duplicate the same piece across categories. AI engines de-duplicate aggressively; you split your own signal.
• Do not write standfirst that repeats your opening paragraph. The duplication splits the citation signal and reads as broken to humans too.
• Do not paste raw text without H2 structure. Flat prose loses to structured prose, every time.
• Do not hide your name or role. AI engines cite named experts far more than anonymous text.

How long until citations?

Crawling: 24–72 hours after publishing, AI engines pull your piece into their index.

First citation: typically 2–6 weeks on a well-defined niche topic. Sometimes faster on a fresh question nobody else has answered.

Compounding: five well-structured pieces in one niche outperform fifty scattered across categories. Pick a topic you can own and ship five times before changing direction. Every new piece on the same niche strengthens the previous ones.

Frequently asked