AI Search · The Spark
How to get cited by ChatGPT, Perplexity and Google AI
A citation puts your name in the answer itself. The seven things that decide whether an AI search engine quotes you, in the order worth doing them.
To get cited by an AI search engine you have to make a clean, authoritative answer easy to extract, then keep it current enough to be trusted. That is the whole game, and it is more mechanical than most people expect.
Our GEO explainer covers what generative engine optimisation is and why it matters. This is the other half: the seven things to actually do, in the order we do them.
Why a citation beats a ranking
A ranking puts you in a crowded room. A citation gets the host to introduce you by name.
When an assistant cites you, it hands your credibility straight to the person asking. There are no competing links beside you and no need for them to click anything. Your name is attached to the answer, which is a fundamentally different position to being result number four.
It is also still wide open. Most businesses have not started, and the platforms overlap far less than people assume, so the same work can win you ground on three surfaces at once.
How the engines decide
The important shift: they are not ranking pages, they are extracting answers. Everything below follows from that one distinction.
The seven steps
One, answer first and explain second. Most business writing follows the university essay: context, background, build up, answer somewhere in paragraph four. An engine will not wait for that and neither will a reader. Make your heading a question, answer it completely in the first two sentences, then use the rest to prove and expand. Put a short bold summary directly beneath every H2 so there is always a clean block to lift.
Two, structure for extraction. Keep paragraphs to two or three lines. H2 for the section, H3 for the sub step. Numbered lists for processes, bullets for criteria, bold for anything that could stand alone as a claim. The test we use on every section: if an assistant pulled this paragraph on its own, would it still make sense? If not, rewrite it.
Three, publish something only you have. This is the biggest separator and the one most people skip. If every number on your page came from somewhere else, you are a secondary source and the engine will cite the original instead. You do not need a research budget. A survey of fifty customers, or an honest analysis of your own campaign data, produces numbers no competitor can copy. Put them in a structured comparison rather than buried in a sentence, give them a date, and say how you got them.
Four, add the schema. Most businesses skip this because it sounds technical. It is a few lines of JSON-LD in the head of the page telling the engine what this is, who wrote it, when it changed and which questions it answers. Article, FAQPage, HowTo and Organisation cover almost everything. It is the cheapest work on this list and the most commonly missing.
Five, keep it current. Assistants are built to give current answers, so they lean hard on recency, and they frequently append the current year to their own sub queries when they go looking. A page you have not touched in a year is quietly losing the citations it had. Carry a visible last updated date, refresh statistics as new ones land, and add genuinely new sections rather than changing a date and hoping. The engines can tell the difference.
Six, work more than one platform. The major assistants draw from noticeably different pools, so being cited by one is a poor predictor of the others. Google's overviews lean on pages that already rank, which means the traditional foundation still earns its keep. ChatGPT rewards depth and completeness. Perplexity rewards directness and visible author credentials. Same content, three different emphases.
Seven, check you are not blocking them. A surprising number of sites quietly exclude GPTBot, PerplexityBot and ClaudeBot in robots.txt, usually copied in years ago or added out of caution about content being taken. The effect is total: you cannot be cited by a system that is not allowed to read you. Unless you have a specific reason, such as material behind a paywall, let them in. Make it a decision rather than an inheritance.
Where to start
- Baseline honestly. Ask ChatGPT, Perplexity and Google the questions your buyers actually ask, and write down who gets named. It is usually uncomfortable and always useful.
- Fix your top twenty pages, not all of them. Answer capsules, heading structure, schema, dates. A week of work on the pages that matter beats a month spread thin.
- Ship one original thing. One survey, one honest analysis of your own numbers. This is what converts a tidy page into a cited one.
- Recheck monthly. Citations move. Track which pages hold them and which are slipping, and treat a slip as a signal to refresh.
None of this is exotic. It is answer first writing, clean structure, something original, correct markup and a habit of going back. The reason it works is that most of your competitors will read a list like this, agree with all of it, and do none of it.
Frequently asked
Questions buyers ask about this
How do I get my business cited by ChatGPT or Perplexity?
Make a clean answer easy to extract and keep it current. In practice that means answering the question directly in the first two sentences of a section, structuring the page with clear headings and short paragraphs, publishing at least some data that is genuinely your own, adding Article and FAQ schema, keeping the page updated, and confirming you are not blocking AI crawlers in robots.txt.
How long does it take to start getting AI citations?
Structural changes such as answer capsules, cleaner headings and schema can shift things within roughly four to eight weeks. Google's AI Overviews usually take longer because they lean on traditional ranking signals, often three to six months. ChatGPT and Perplexity can pick up well structured, authoritative content noticeably faster.
Do I need original research to be cited?
It is the single biggest differentiator. If every figure on your page is borrowed, the engine will generally cite whoever published it first. You do not need a large study though. A survey of fifty customers or a careful analysis of your own project data gives you something citable that no competitor can replicate.
Should I block AI crawlers in robots.txt?
Only with a specific reason, such as paywalled material. Blocking GPTBot, PerplexityBot and ClaudeBot removes you from the answer layer entirely, and many sites are doing it accidentally through rules copied in years ago. It is worth checking your robots.txt today and making it a deliberate choice.
Does being cited by AI actually drive business?
The traffic volume is smaller than organic search but the quality is different. Someone arriving from a citation has already been told by a system they trust that you are a credible source, so they arrive pre-qualified rather than comparison shopping. That tends to show up as shorter cycles and better conversion rather than raw volume.
Working on a real engine? Start with a conversation.
Tell us where you are. We will tell you what we see and where we would start.