Network of circuits representing AI search engines
    ← Back to Blog

    May 2026 · Teralith · 9 min read

    The New SEO: Getting Cited by ChatGPT, Perplexity, and Google AI Overviews

    Your website is invisible to ChatGPT. So is the website that won page-one rankings two years ago. So is most of the internet, actually — and that is a problem if you depend on search to bring in customers.

    Something has changed about how people find businesses online, and it has happened fast. In 2024, a Google search returned ten blue links and a few ads. In 2026, the same query is just as likely to surface a single AI-written summary at the top of the page — Google AI Overviews — that answers the question outright. Many users never scroll past it. The same is true on Bing, where Copilot answers replace traditional results entirely; on ChatGPT and Claude, where users now search the live web inside the chat window; and on Perplexity, which is built around the idea that you should never have to click again.

    This is the shift from SEO to AEO — Answer Engine Optimization. And it is not a future thing. It is happening on your site right now, whether you have prepared for it or not.

    The Click That Never Was

    The traditional search funnel was simple: someone typed a query, Google ranked the relevant pages, the user clicked one, and landed on your site. You measured success in clicks and conversions.

    The new funnel is messier. Someone types a query. The search engine — or the AI chatbot — synthesises an answer from multiple sources. Your site might be cited as one of those sources, with a small link or a brand mention. Most users get the answer they wanted and leave. A small fraction click through to verify, learn more, or buy.

    The first thing this changes is your traffic numbers. Industry research suggests informational queries are now seeing zero-click rates of 60% or higher in some categories. The clicks you used to get for "how much does a website cost" or "what is AI automation" are increasingly being absorbed by an AI summary at the top of the page.

    The second thing it changes is what counts as a win. If your business is mentioned in the AI's answer, you have done something almost as valuable as ranking on page one — even if no one clicks. You have placed your brand in front of a buyer at the exact moment they are forming a decision, and you have done it with the implicit endorsement of the search engine.

    That is the new game. Not "how do I rank?" but "how do I get cited?"

    What AI Search Engines Actually Look For

    The major answer engines — Google's AI Overviews, Microsoft Copilot, OpenAI's ChatGPT search, Anthropic's Claude search, Perplexity, You.com — do not all use identical ranking signals, but the pattern across them is consistent. To get cited, you need to look like a source the model can trust and a source the model can cleanly extract from. That breaks down to about six things.

    1. Domain authority and link signals. Same as classic SEO. The AI engines were trained on the open web, and the open web's signals of trust are still backlinks, brand mentions, and topical depth. A new domain with no inbound links does not get cited even if its content is excellent.

    2. Structured data. Schema.org markup — Article, FAQPage, Service, Organization, BreadcrumbList — is how you label your content for machines. AI crawlers use it to extract facts cleanly. A page with no structured data is harder to parse and easier to skip.

    3. Declarative, fact-first writing. The models prefer prose that states facts cleanly: short paragraphs, a single idea per paragraph, lists where they fit, definitions up front, claims followed by evidence. Florid copy and clever framings are harder to extract. If you cannot pull a quotable sentence out of a paragraph, the model probably cannot either.

    4. Specific answers to specific questions. A page titled "About Our Services" gets cited rarely. A page titled "How Much Does a Custom E-Commerce Website Cost in 2026?" gets cited more often, because it lines up with a query someone actually types. Build pages around questions, not categories.

    5. Recency. Models discount old content for time-sensitive topics. A pricing article from 2022 will lose to a pricing article from 2026 even if the older one is technically better. Date your content. Update it. Republish when it changes.

    6. Topical authority. A site with one article on AI automation looks shallow. A site with twelve articles covering different angles of AI automation looks like an authority. The models cite authorities. Build clusters, not one-offs.

    None of this is conceptually new. What is new is how unforgiving the engines have become about it. Vague pages, empty schema, and no recent updates used to mean ranking on page three. Now they mean not being seen at all.

    The Playbook: What to Do This Quarter

    You do not need a six-month project to start showing up in AI answers. Most of the work is small, deliberate, and compounds. Here is the order I would prioritise it in for a small or mid-sized business.

    Audit your structured data. Open one of your service pages in Google's Rich Results Test. If it returns nothing, you have no structured data, and that is the cheapest thing to fix. At minimum every page should declare what it is — Service, Article, Product, Organization — and how it relates to your business. FAQPage on any page with a Q&A section is an easy win.

    Rewrite your most important page in declarative prose. Pick the single page you most want cited — usually the home page, your highest-margin service, or your strongest blog post — and rewrite it so the first sentence of every paragraph is a complete, quotable claim. Read it out loud. If a sentence sounds like marketing fluff, rewrite it as a fact.

    Add a real FAQ. Five to ten questions per major service, written exactly as a customer would search for them. Real questions, real answers. Mark up the section with FAQPage schema. This is the single most-cited content type across AI summaries because it lines up perfectly with how people query.

    Date your content and update it. Add an "updated" date to every long-form page. Set a quarterly calendar reminder to review each one. Even a small update — pricing, a new statistic, a fresh example — refreshes the page's recency signal. Old content does not have to die; it has to be tended.

    Publish on a beat. One blog post a month is enough to move the needle on topical authority for most niches. Three is better. The compounding effect is real — the tenth post on a single topic gets you cited far more often than the first. Build the cluster.

    Don't block AI crawlers. This is the one most agencies and security platforms get wrong. There is a school of thought that says AI scrapers are stealing your content, so block them at the firewall. For a marketing-led business, this is self-sabotage. The whole point is to be cited; you cannot be cited if you cannot be read. Cloudflare's "AI Labyrinth" feature, for example, is designed to feed crawlers fake content. Useful for some publishers; actively harmful for almost every services business that wants AI engines to recommend them.

    How to Measure What's Actually Happening

    The metrics most analytics tools report — sessions, pageviews, organic clicks — measure clicks. Citations are not clicks. So your traditional dashboards will quietly understate what is going on.

    A few practical measurements:

    Manually query your own brand. Once a month, type your company name and your top services into ChatGPT, Perplexity, Google AI Overviews, and Claude search. Are you cited? When? In what context? Make a screenshot log. Watch for changes month over month.

    Track branded search. If your AEO is working, more people will hear your name in an AI answer and then search for your brand directly. Branded search volume is a proxy for citation reach. Watch it climb in Google Search Console.

    Watch zero-click queries in GSC. Search Console reports impressions and clicks per query. A query with high impressions and falling clicks is one where AI Overviews are absorbing the result. That is not a problem if you are cited — it is the new normal.

    Measure conversion paths, not first touches. A user who first heard your name in a Perplexity answer, then searched, then booked, will look in your analytics like a "direct" or "branded organic" lead. It came from AEO; the dashboard will not say so. Long-window attribution and post-purchase surveys are how you actually catch this.

    What This Doesn't Replace

    Traditional SEO has not died. It has folded into AEO. Every signal that helps you rank — fast pages, mobile-first design, clean URLs, internal linking, real backlinks — also helps you get cited, because those are the same signals the AI engines use to decide what content is trustworthy. If you are reading this and worrying that years of SEO work is now obsolete, it isn't. It is the foundation. AEO is the storey on top.

    What has died, or at least has changed substantially, is the assumption that ranking equals success. In 2026, getting cited equals success. Sometimes a citation comes with a click and sometimes it doesn't — but the citation itself is the win, because it places your brand in front of a high-intent buyer at the exact moment a decision is being made, with the trust transfer of the search engine that put you there.

    Where to Start Tomorrow Morning

    If you do nothing else: write one new long-form page on the most specific question your buyers ask, mark it up with Article and FAQPage schema, and publish it. That is the AEO equivalent of planting a tree. Every month after that, the citation possibilities compound.

    We do this for clients as part of our SEO and AI Automation services — the structural work, the schema, the content cadence, the measurement. If you would like a free audit of how visible your business currently is in AI search, book a 30-minute call and we will tell you exactly where you stand and what to do next.