If your SEO plan still starts with "rank higher, get more traffic", it is a plan for 2019, and the numbers say so.
Your customers now get a first answer from an AI Overview before they see a single link.
They ask ChatGPT which suppliers to shortlist. They let an assistant do the comparison before they open a website.
Google says AI Overviews reach more than two billion people a month and that AI Mode passed a billion monthly users within its first year.
That is the shape of search going into 2027, and it changes how a business gets found, how it earns trust, and where the enquiry actually comes from.
Here is what I would be doing about it now, with the evidence for each step.
SEO and AEO are two jobs, and you need both
SEO is the work of making your site visible in Google and Bing. AEO, answer engine optimisation, is the work of making your content easy for an AI system to understand, lift and cite when it answers a question.
They overlap. A site that does SEO well has the foundation for AEO. But the second job puts far more weight on how clearly you answer a question, and on whether the engine can take one clean paragraph from your page and quote it.
By 2027 that second job deserves its own line in the plan, because a ranking without a citation is increasingly a ranking nobody sees.
1. Getting found is no longer getting clicked
SparkToro's 2026 study of US Google searches found that 68% ended without a click, up from 60% in 2024. Pew Research tracked 900 adults across nearly 69,000 searches and found people clicked a result 8% of the time when an AI Overview was present, against 15% without one. Only 1% clicked a link inside the Overview.
Worse for your reports, an AI assistant can recommend your business during someone's research and that recommendation never appears in your analytics at all.
So organic sessions alone no longer tell you whether you are visible. A traffic drop may mean your SEO has failed, or it may mean customers are seeing your answer somewhere you are not measuring.
What to do: track impressions, enquiries and conversions alongside sessions in Search Console and GA4. Then run the questions your business should be the answer to through ChatGPT, Gemini, Perplexity and Google's AI features, and record who gets named.
I set out what to measure when nobody visits in detail, including why one check on one day is not a measurement.
2. Make your expertise impossible to skip
AI can write an answer on anything. It cannot write your answer, the one built on cases you have handled, prices you actually charge and outcomes you can show.
Seer Interactive's dataset of 2.43 billion organic impressions found that when an AI Overview appears, brands cited inside it earn 120% more clicks per impression than brands that are not. Uncited brands saw their click-through rate fall 67% across 2025. The citation is where the remaining traffic lives, and the engines cite sources that say something specific.
Google's E-E-A-T framework, experience, expertise, authoritativeness and trust, is the description of what gets cited, and there is no score to game.
It comes down to whether a named person with real credentials wrote the page, whether the page shows the work behind its claims, and whether the information is current.
What to do: put credible people behind the content, with bios and credentials on the page. Link to your own case studies and data. Keep dates and figures current.
Build pages that connect, so a visitor or a crawler can go from one answer to the next instead of landing on an orphaned post. I showed what E-E-A-T looks like on a small firm's page, with the before and after.
3. Write so an answer engine can lift the answer
An AI system has to parse your page before it can use it. The easier that is, the better your chance of being the source. That does not mean keyword stuffing or two hundred AI-written FAQs. It means being plain about what you do, who you help and what they need to know.
If someone asks "how much does commercial solar installation cost", a page that explains the factors behind the price gets cited. A page that promises competitive rates does not.
Citation is a passage-level decision. The engine does not cite your page; it cites one paragraph of it. So every heading should be a question a customer would ask, and the first sentence under it should answer that question in words that stand alone when quoted.
I wrote up why answer engine optimisation is a passage-level job with examples of paragraphs that get lifted and paragraphs that get skipped.
What to do: review your most important pages against the questions customers ask before they buy. Use descriptive headings, short direct answers, and a logical order. Check your structured data is accurate and matches what visitors see.
It helps engines understand the page. It does not buy citations, and anyone who tells you otherwise has not tested it.
4. Prepare for agents, not just answers
The next step is agentic AI: tools that act on the answer. A buyer asks an assistant to compare three suppliers, check their services, and recommend one. To survive that process, your business has to be represented accurately everywhere the agent looks, not just on your homepage.
Nobody can honestly say how this will work on every platform by 2027. What is already clear is that the agent reads your pricing page, your reviews, your directory listings and your contact details, and if they disagree with each other, it moves on to a supplier whose details agree.
What to do: make services, locations, prices and contact details consistent across your website, your Google Business Profile, your social profiles and the review sites. Make the important pages crawlable and the site usable on a phone. I covered what AI crawlers can and cannot read, and the list of what blocks them is longer than most owners expect.
5. Build the reputation the engines actually read
Your website is one source among many, and for AI engines it is often not the first.
Surfer SEO's analysis of 46 million AI Overview citations found YouTube took 23% and Wikipedia 18%. Reddit takes around a fifth in several studies. Company websites account for about 31%, the largest single category, but they are competing with video, forums and reference sites for the slot.
In mid-2024 around three quarters of AI Overview citations also sat in the organic top ten. By early 2026 that overlap had fallen to between 17% and 38%, depending on the study. The engines are choosing sources that do not rank.
That is where digital PR earns its place again. Coverage in a trade publication, a presentation someone wrote up, an industry body's directory, a well-answered question on a forum: all of it is readable by the engines and all of it feeds the citation. I explained why those third-party mentions now beat your own site for AI visibility.
What to do: publish original research and opinions worth quoting. Contribute to the publications your customers read. Watch how your business is described online and correct what is wrong. Do not buy links or place guest posts purely to move rankings. Neither builds the kind of mention an engine trusts.
What to prioritise before 2027
You do not need to tear up your SEO. The fundamentals have not changed; the ceiling on certain content has, and the measurement has.
Start with the pages that bring in enquiries. Update anything out of date, put a named expert behind each one, and rewrite the opening of every section so it answers the question in the first sentence. Make your details consistent everywhere the engines look.
Then set up a way to see whether you are being named, because by 2027 that will matter more than where you rank.
Google's own guidance on AI features says there is nothing special to do for AI search beyond publishing content people would want to cite. That is the whole job, and it was always the job. The difference now is that the businesses that skip it do not just rank lower. They disappear from the answer.
If you want to know whether the engines name your business today, and who they name instead, that is the first thing we check.
