How to Get Cited in ChatGPT AI Overviews (2024 Guide)

Here is the uncomfortable bit nobody wants to put on the homepage. Ranking number one no longer guarantees anyone reads your page. A robot reads it, summarises it, and hands the answer to your reader before they ever see your blue link. So the new game is not just ranking. It is getting cited in ChatGPT and AI Overviews so your brand is the source the machine trusts.

We are SEO Engico, a London agency run by two ex-mechanical engineers who treat marketing like a system. So this guide is not vibes. It is what we actually see in client campaigns, backed by data we have checked ourselves. Let's get into what works in 2026, what is a waste of your weekend, and where most brands are quietly losing.

Why getting cited matters more than it did last year

The click economy is shrinking, and we have the receipts. In July 2025 the Pew Research Center tracked the real browsing behaviour of 900 US adults across nearly 69,000 Google searches. When an AI summary appeared, people clicked a normal search result only 8% of the time, against 15% when there was no summary. Roughly half the clicks, gone.

And the links inside the AI summary itself? People clicked those in just 1% of visits. Read that again. The box answers the question, the reader nods, and the session ends. Pew found people abandoned the page entirely 26% of the time when a summary showed up, versus 16% without one.

So if you are still measuring success purely by sessions, you are grading yourself on the wrong exam. The win now is being named inside the answer. Even when nobody clicks, a citation puts your brand in front of the buyer at the exact moment they are deciding. That is brand authority you cannot buy with a banner ad.

Our take: citations are the new featured snippet, except they compound. Once a model learns to associate your brand with a topic, it keeps pulling you in across thousands of related prompts. We dig into the mechanics of this in our breakdown of how to get your brand into AI answers.

What is the difference between getting cited and ranking?

Ranking means Google decides your page deserves a spot on a results page. Getting cited means a language model decides your page is worth quoting inside its generated answer. Related, but not the same job.

The two used to be almost interchangeable. Not anymore. Ahrefs analysed 863,000 keyword SERPs and four million AI Overview URLs in March 2026 and found only 37.9% of pages cited in AI Overviews actually rank in Google's top ten for that query. In their July 2025 study that number was around 76%. So in roughly eight months, the overlap nearly halved.

Where did the rest go? Ahrefs found citations now pull heavily from pages ranking 11 to 100 and even beyond position 100. Google's system runs what it calls query fan-out, spinning your one question into a cluster of related searches and gathering sources across all of them. Page two is no longer the place where dreams go to die.

This is genuinely good news if you are not a household name. You do not need to win the head term to get quoted. You need to be the best answer to a specific slice of the question. More on that below.

How do ChatGPT, AI Overviews and Perplexity each pick their sources?

Quick answer: they do not share a rulebook. Profound studied 680 million citations from August 2024 to June 2025 and the headline is that there is no universal top source. Each engine has its own taste, and pretending otherwise is how brands waste budget.

Here is the short version of what that data shows, plus what we see in our own tracking.

• ChatGPT leans on authority. Wikipedia was its single most cited source at 7.8% of citations in the Profound data. It likes encyclopaedic, structured, well established sources. If your topic touches a real entity, having a clean Wikipedia presence quietly helps, which is why we keep banging on about how Wikipedia shapes brand SEO and LLM citations.
• Perplexity loves the forum floor. Reddit was its leading source at 6.6%, and Wikipedia did not even make its top sources. Perplexity wants recent, opinionated, real-person content with explicit links. Tidy and dull does not cut it there.
• Google AI Overviews sit in the middle. Reddit led at 2.2% and Wikipedia trailed at 0.6%, but Overviews draw from a much wider, more diversified pool tied to its own search index. They reward content that already earns Google's trust the old fashioned way.

The practical lesson we hammer with clients: do not optimise for "AI" as if it is one customer. You are pitching three different editors with three different briefs. We map exactly how to play each one in our guide to ChatGPT search optimisation strategies.

Front-load the answer, then earn the read

Models are skim readers with a deadline. They lift the answer from wherever it sits cleanly near the top, then move on. So burying your best line in paragraph fourteen is how you stay uncited forever.

Lead with a direct, quotable answer. Then expand with the evidence, the nuance and the examples. This is the same instinct behind a good featured snippet, just turned up to eleven. We pulled the data together in our look at why ChatGPT favours citations from the first 500 words, and the pattern holds across engines.

Here is the structure we use on client pages built to get cited:

1. Open with a one or two sentence answer to the exact question in the heading. No throat-clearing, no "in this article we will explore".
2. Follow immediately with the proof. A stat, a named source, a concrete example.
3. Then go deep. Edge cases, counterarguments, the bits the top result skipped.
4. Use a clear question as the heading, because that is how people prompt and how models match.

What we actually see: the pages that get cited the most in our campaigns are not the longest. They are the clearest. A 1,400 word page that answers one question brilliantly beats a 4,000 word page that rambles through five.

Structure your content so a model can lift it cleanly

Language models are pattern machines. Give them clean patterns and you make their job easy, which makes a citation more likely. Give them a wall of text and they go and quote someone tidier.

That means descriptive headings phrased as questions, short paragraphs, lists where lists make sense, and one idea per section. It also means being explicit. State the fact, attribute it, date it. A sentence like "according to Pew Research in July 2025, clicks roughly halved" is far easier for a model to trust and reuse than a vague "studies show clicks are down".

Now, a myth we need to kill. You do not need a secret AI file or magic markup. Google's own documentation is blunt about this. The official Search Central guidance on AI features says there are "no additional requirements to appear in AI Overviews or AI Mode," and that "you don't need to create new machine readable files, AI text files, or markup," including "no special schema.org structured data that you need to add."

So if a tool is selling you an AI-only optimisation hack, be suspicious. We dug into one of the most hyped examples and shared the data in our analysis of llms.txt and whether it moves citations. Spoiler: the fundamentals matter more than the gimmick.

The big secret: it is still SEO

The most reassuring sentence in this entire field comes straight from Google. In its guide to optimising for generative AI search, Google states plainly that "from Google Search's perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO."

So all the GEO and AEO acronyms people are inventing? Useful framing, but the underlying work is the same craft you already know. Google's guide spells out the priorities: bring a "unique point of view" rather than recycling common knowledge, keep content "helpful, reliable, and people-first", stay crawlable because the models use "publicly accessible, crawlable content", and support text with strong images and video.

We have said this to clients who arrive convinced they need a brand new "AI strategy". Most of the time they need their existing technical and content SEO done properly first. Crawlability, page experience, genuine expertise. Get those right and AI visibility tends to follow, because the same signals feed both systems.

The flip side: the brands that fail at AI search usually fail at the basics. We pulled apart the common patterns in our piece on why brands fail at generative engine optimisation, and "we skipped the fundamentals and chased hacks" is the recurring theme.

How small sites win without huge authority

You do not need to be a giant to get cited. In fact, smaller, sharper sites have a real opening right now, and the Ahrefs query fan-out data proves it. When citations are pulling from pages ranking 11 to 100 and beyond, depth on a narrow topic beats brand size.

Here is the move we coach clients through.

• Go absurdly specific. Do not write "email marketing tips". Write the definitive answer to "how to win back SaaS trial users who signed up but never activated". Big brands ignore that long tail. You can own it outright.
• Build clusters, not orphans. One brilliant page is good. A connected set of pages around one topic signals real depth, and models reward that web of context. Internal links are doing actual work here, not just passing link equity.
• Earn second-order citations. Get quoted in sources the models already trust. Contribute real data to an industry report, write expert commentary, get into a credible roundup. When a trusted source cites you, the engine starts associating your brand with that expertise. This is where smart link building and digital PR earn their keep, and it overlaps neatly with our white-label link building work for agencies.
• Be the freshest answer. Recency carries weight, especially on Perplexity and in fast-moving topics. Stale content gets quietly demoted in favour of whoever updated last.

Our honest opinion: the underdog play is the best play for most businesses. Trying to out-authority Wikipedia is a losing battle. Being the single best answer to a question nobody else bothered to answer well is very winnable.

How big is this getting? The volatility nobody warns you about

Worth a reality check. AI Overviews are not a steady, predictable line going up. Semrush tracked over 10 million keywords through 2025 and found AI Overview coverage started at 6.49% of queries in January, peaked at 24.61% in July, then settled back to 15.69% by November. It went up, then came down, then steadied.

The same volatility shows up in which sources get cited. Citation shares for individual domains can swing wildly month to month as the platforms tweak their systems. That is exactly why a "set it and forget it" approach fails here.

So treat AI visibility as a living metric, not a project you finish. We bake quarterly refreshes into client retainers for this reason. Update the stats, add the new examples, re-check the sources, and you keep signalling that your content is current and maintained, which is precisely what these systems favour.

How do you actually track AI citations?

Honest answer: the tooling is still catching up, and a lot of it is manual graft. There is no perfect dashboard yet, so here is the workflow we use.

1. Prompt the engines yourself. Run the questions your customers ask across ChatGPT, Google AI Overviews and Perplexity. Log when you appear, what content gets pulled, and which competitor keeps showing up instead of you.
2. Watch Search Console for the tell. Impressions climbing while clicks flatline on informational queries is a classic AI Overview fingerprint. You are being shown, just not clicked.
3. Track brand mentions, not just links. Engines increasingly mention brands without linking them. A name-check still shapes buyer perception, so count it.
4. Re-test after every platform wobble. Citation shares move fast. What got you quoted in February might not in May.

If running all that across dozens of prompts every month sounds like a part-time job, that is because it is. It is exactly the kind of measurement loop we run inside our AI search visibility service, where we treat citations as a tracked metric with a target, not a hopeful guess.

The plan we would actually run

If you handed us a site tomorrow and said "get us cited", here is the order we would work in. No fluff, just the sequence.

1. Fix the fundamentals first. Crawlability, indexing, page speed, genuine helpful content. Google says AI search is still SEO, so we start where the leverage is.
2. Map the real questions buyers ask and write the single best answer to each, front-loaded and clearly structured.
3. Build the topic cluster so the pages support each other and signal depth.
4. Earn trusted external mentions through data, PR and link building so the engines learn to associate your brand with the topic.
5. Track citations across all three major engines monthly, then refresh the winners and rework the losers.

None of this is magic. It is disciplined SEO pointed at a slightly new target. The brands getting cited in 2026 are not the ones chasing the latest acronym. They are the ones answering real questions better than anyone else and keeping it current.

If you would rather not run the experiments and the monthly prompt-tracking yourself, that is literally our job. Tell us what you sell and who you sell to, and we will tell you where you are getting left out of the answers and what it would take to fix it.