How to Get Cited by ChatGPT and Perplexity

The single most important fact, and it is a technical one: ChatGPT search retrieves from Bing's index.

Which means a site that is not in Bing is invisible to ChatGPT, regardless of how it ranks in Google. And almost nobody has set up Bing Webmaster Tools. That is step one, it is free, it takes twenty minutes, and it is the highest-leverage thing on this page.

Perplexity runs its own crawler and index, cites sources prominently, and sends genuine referral traffic, which makes it the most rewarding of the platforms to be visible in and the easiest to measure.

After that, both reward the same things: rank in the underlying index, be crawlable by the retrieval bot, answer the question in the first 90 words, write claims that survive being lifted out of the page, and be a named, credentialed, verifiable entity.

ChatGPT: the Bing problem

ChatGPT search is Bing-backed.

So the funnel is: be in Bing's index → rank in Bing → be crawlable by OAI-SearchBot → be extractable.

And the first step is the one everybody has skipped.

Bing Webmaster Tools is free. Submit the sitemap. You can import directly from Google Search Console, which takes about five minutes.

Then check that OAI-SearchBot is not blocked, in robots.txt and at the CDN. → The AI crawler guide

And then use the thing nobody knows about: since June 2026, Bing Webmaster Tools has provided four AI-visibility metrics, free, in preview. First-party data about your visibility in AI answers, from the index that powers the largest AI assistant in the world, at no cost.

The paid AI visibility tools are charging for a worse version of this.

Perplexity: its own index, and the one that pays you back

Perplexity crawls independently with PerplexityBot and maintains its own index.

And it is the most generous of the platforms: it cites prominently, it links, and it produces measurable referral traffic in your analytics, which means you can actually prove the value to a client. The others largely cannot be proved this way, and that is a genuine reporting problem.

What Perplexity favours:

Recency. It leans hard on fresh content. A page updated last month beats an equivalent page from 2023.

Direct answers. It is a question-answering engine, and it rewards content built as answers.

Sources it can verify. Named authors, cited data, primary sources.

And it is noticeably kinder to smaller sites than Google is, which is the most interesting thing about it. A genuinely good, specific, well-structured page from a small domain can be cited alongside a national publisher.

That is a real opportunity and it will not last forever.

What both reward

Answer in the first 90 words. Under the H1. Before the preamble.

Self-contained claims. Every sentence must survive being lifted out. "As discussed above" is unquotable, and a great deal of professional content is written entirely in that register.

Question-form headings. The heading is the retrieval hook.

Tables and lists. Pre-structured content is easier to extract, and easier to extract means more likely to be quoted.

Specificity. A model quotes a number, a date, a named mechanism. It does not quote "best-in-class." "$271.49 cost-per-click on divorce lawyer terms" gets quoted. "Legal SEO is very competitive" does not.

Named authors with Person schema and a sameAs graph. Anonymous content does not get cited, because the model has no basis to attribute it.

Content out of accordions. The single most common silent failure.

Being in the index in the first place, which is the 80% everybody skips past to get to the tactics.

→ What is answer engine optimisation?

The thing that matters more than any tactic

Being an entity.

When somebody asks a model "who is the best white-label SEO provider for agencies," the model is not running a search and ranking results. It is drawing on an association it has built between a concept and a set of names.

That association was built from your name appearing, repeatedly, alongside your category, across the web, in other people's writing.

Which means the work that gets you named is: coverage, mentions, citations, being quoted, being listed, being on the professional register, being in the industry directory, being the person quoted in the trade publication.

And notice that a language model does not check the rel attribute. A nofollow mention builds the entity exactly as well as a dofollow one. → Do nofollow links help SEO? · Topical authority and entity SEO

This is slow, unglamorous, and it is the actual answer.

And what you cannot do

You cannot pay to be cited. There is no ad slot in an AI Overview citation.

You cannot game it with keyword density, and anybody selling you "LLM keyword optimisation" is selling you 2011 with a new noun.

You cannot guarantee it. Citations are volatile. The same query can produce different sources on different days, and a client watching daily will be miserable.

And you cannot measure it with one number, because roughly 91% of URLs cited by language models appear in only one platform. → The best LLM SEO tracking tools