Google Killed Classic SEO at I/O 2026. Here Is What Replaced It.

Every era of local marketing has had a directory. The phone book gave way to Google. In May 2026, on the stage at I/O, Google made it clear that Google is giving way to something else, and it is building that something itself.

This is not a doom post. Our clients get more leads today than they did a year ago. But the mechanism changed, and most agencies are still selling the old one.

What Google actually announced

Strip away the keynote language and four things matter for anyone who sells local services.

AI Mode reached a billion users. A search that used to return ten links now returns one composed answer, with a handful of sources cited inside it. Google said AI Overviews reach 2.5 billion people every month. That is not an experiment anymore. It is the default surface.

Information Agents run around the clock. Google described agents that research, compare, and take action on the user's behalf without the user browsing a single website. A homeowner with a leak does not open six roofing sites. She asks, and something answers with two or three names.

Generative UI builds the page on the fly. Instead of sending the user to your site, Gemini can assemble a widget, a comparison table, or a booking flow inside the answer, hydrated from structured data it found on the web. Your website is becoming a data source, not a destination.

Ads are moving inside the answer. The day a business can pay to be the recommendation, the honest question for every owner becomes "was I recommended, or was I bought?" We are already tagging that distinction in our own tracking, because the difference will matter in every report.

Why "rank first" stopped being the finish line

SISTRIX measured what happens to the number one organic result when an AI Overview renders above it: click-through fell from about 27 percent to about 11 percent. Same position. Same page. Less than half the clicks.

That is the whole story in one number. Ranking first still matters, but it is now a qualifier, not the prize. The prize is being one of the sources the answer cites. Digital Applied found brands cited inside AI Overviews saw organic clicks lift around 35 percent. The citation is where the traffic went.

For a local service business the consequence is blunt: you can hold the number one spot for "roofer in Chattanooga" and still lose the job to the company the AI named.

What still works, and what does not

Some of the old playbook survives, some of it is now noise.

Still works:

• A fast, crawlable site with facts on it. AI engines cite pages they can read and trust. Vague brochure copy gets skipped. A page that says what you do, where, for whom, and at what range gets quoted.
• Google Business Profile, kept current. For local answers the engines lean hard on the profile, its reviews, its categories, and its Q and A. It is the single most cited local source we see.
• Reviews with substance. A review that says "fixed our flat roof leak in Hixson the same week" teaches the engine something a five star rating does not.
• Structured data that matches reality. Schema is how Generative UI hydrates a widget. If the schema on your page says one service area and the footer says another, you have taught the machine to distrust you.

Does not work anymore:

• Rewriting title tags to chase a keyword.
• Publishing a dozen thin city pages that all say the same thing.
• Reporting impressions as if they were customers.
• Selling "we got you to page one" as the outcome.

The citation playbook for this quarter

This is what we run for every client site now, in order.

1. Check whether AI crawlers can even reach the site. A surprising number of businesses block the retrieval bots that ChatGPT, Perplexity, and Apple use to answer questions, usually by accident. We wrote about one case where a company was invisible to Apple Intelligence and nobody knew.
2. Approve the facts once. Services offered, services not offered, service area, hours, what the business will and will not promise. Get the owner to sign off. Everything downstream reads from that one approved record.
3. Publish the facts as structured data. LocalBusiness, Service, areaServed, hours, and a FAQ that answers the questions customers actually ask, written from the approved facts and nothing else.
4. Measure the answers, not the rankings. Ask ChatGPT, Perplexity, Gemini, and Google AI the questions your customers ask, in your city, on a schedule. Record who gets cited. That is the new rank tracker.
5. Close the loop on every lead. The visitor an AI answer sends you is a person with a problem right now. Answer in seconds, in the business's voice, and offer a real appointment. If that step fails, everything above it was wasted.

The part most agencies will get wrong

The temptation is to treat this as a content problem and start publishing "GEO optimized" articles at scale. We tried the automated version of that on our own fleet and killed it: dozens of generated posts, none worth publishing. AI engines are not looking for more text. They are looking for a source they can trust about a specific business in a specific place.

Trust is not a keyword. It is a business that answers the same way everywhere: on its site, on its profile, in its reviews, and in the reply a customer gets at eleven at night.

That is the platform we are building, and it is why our own site runs on it. If you run an agency and want to see the measurements on a real client, ask us for the early access demo. We will show you the answers the engines are giving in your market today.