After AI Overviews: A Modern SEO Strategy for 2027

Somebody forwarded you a chart this spring showing organic click-through rates falling off a cliff. The headline said SEO was dead, the replies said it wasn't, and you closed the tab without an answer. Here's the answer.

What happened is simpler than the hot takes. Google moved AI Overviews from experiment to furniture, then spent 2026 rebuilding Search around AI Mode. Some queries stopped sending clicks; the ones that pay your bills barely budged.

What actually changed by mid-2026

For years the SERP was ten blue links; now it's an answer with a list of sources underneath. Three things moved at once:

• Adoption. BrightEdge's nine-industry tracker put AI Overviews on about 48% of queries in March 2026, up roughly 58% year over year. Trackers disagree on the exact figure, not the direction; health and education run far above average.
• Click behavior. SparkToro, using Similarweb clickstream data, found 68% of US Google searches ended without a click in the first four months of 2026, up from 60% in 2024. Ahrefs put the CTR hit for top-ranking pages at 58% when an Overview appears, up from 34.5% eight months earlier. Pew's panel said the same thing from the user side: an 8% click rate on searches with an AI summary versus 15% without.
• The interface itself. At I/O in May 2026 Google called its redesigned, AI-native search box the biggest upgrade to Search in more than 25 years, and said AI Mode had passed a billion monthly users. Blue links aren't gone, but they're optional on far more searches.

Where the clicks went

"Clicks dropped" is true and useless. What matters is which clicks dropped. Seer Interactive's panel of 53 brands and 5.47 million queries is the clearest map we've seen: Overviews show up on roughly 95% of comparison queries and 86% of question queries, and on only about 5% of transactional ones. That split explains most of the argument.

In that same Seer data, pages cited in an Overview saw around 2.1% CTR while uncited pages on the same result page saw about 0.9%. A no-Overview SERP still beat both at roughly 3.3%. The Overview costs everyone something, then splits what's left between the cited and the ignored. There's a floor under it: Overview CTR bottomed out at 1.3% in December 2025 and had recovered to 2.4% by February 2026.

Ranking used to be the finish line. Now it's the qualifying round, and the race is for the citation.

The five-layer 2027 SEO stack

Every page should hit all five. Skip one and you're optimizing for a SERP that no longer exists.

Layer 1: Answer-first content

The first 100 words of a page should answer the question the title implies. Not "in this article we'll explore." The answer, then the depth. Two reasons: a human who arrived from an Overview already read the shallow version, and the next AI engine to crawl your page wants an extractable answer, not filler. That is all answer engine optimization really is.

Layer 2: Structured data (schema), for the right reasons

This is where the GEO pitch decks lie. In May 2026 Google published its guide to optimizing for generative AI features, and the mythbusting section is blunt: you don't need an llms.txt file, you don't need to chop your pages into AI-friendly chunks, and there is no special schema.org markup that gets you into AI features. From Search's perspective, AEO and GEO are still SEO.

So don't add schema expecting an AI citation to fall out. Add it because it still earns what it always earned: rich results, local pack visibility, product and review displays.

Layer 3: E-E-A-T signals (real authors, real reviews)

Experience, expertise, authoritativeness, trustworthiness. 2026's core updates made ignoring it expensive. The March update was brutal: roughly 79.5% of top-three URLs changed position and about a quarter of top-10 pages dropped out of the top 100 entirely. Gainers skewed toward official and institutional domains, established brands, and specialist platforms, while aggregators, directories, and comparison sites lost ground. A second core update landed in May.

Practically: name your authors, write real bios, mark them up with Person schema, and let customer reviews show on the page instead of in a third-party widget.

Layer 4: Distribution & citation worthiness

AI engines cite sources they've already encountered, and they lean hard on a small set: Pew found Wikipedia, YouTube, and Reddit alone accounted for about 15% of the links inside AI summaries, with .gov sources appearing three times as often as they do in standard results. Institutional trust is the currency, and link farms don't buy it. What works is showing up where your industry talks: trade publications, Reddit threads, transcribed podcasts, partner sites, local directories. Boring, slow, durable.

Layer 5: Technical foundation (speed, accessibility, indexing)

None of it matters if your page won't load or won't render. Core Web Vitals, accessibility, clean indexing, server-rendered content where it counts: that's the floor. Google's own AI-features guidance puts crawlability and indexing at the top of its list — unglamorous, and also why plenty of sites are invisible. If your content only appears after a click, a scroll handler, or a client-side fetch, assume the thing summarizing your industry never saw it, and that goes double for agent-driven browser visits.

What to stop doing before 2027

Four tactics that went from diminishing returns to harmful:

• Keyword stuffing in any form, including the "natural" version where a tool told you to say the phrase eight times. Ranking systems read meaning, not phrase density.
• Thin aggregator and listicle pages. "Top 10 X" with two-sentence entries. The Overview eats these whole, and the March 2026 core update went after exactly this shape of site.
• AI-spun content at scale. Google's scaled content abuse policy is explicit that it doesn't care how the pages were produced: unoriginal pages published at volume to catch rankings are the violation. A global spam update shipped in June 2026, and Google has confirmed the spam policies apply to AI Overviews and AI Mode too. Automation isn't the problem — we automate this blog — the editorial standard is.
• Buying "GEO" packages built on myths. If the deliverable is an llms.txt file or chunking your content for the models, you're paying for work Google has publicly said does nothing.

How to measure success when "rank" means less

A single position number is a worse proxy than it used to be: a page can rank #2 and get no clicks because the Overview answered the query. Track these instead:

• Branded search volume. Are more people typing your company name? That's the canary for everything else working.
• Citation share. Is your domain showing up in AI Overview source lists, Perplexity answers, ChatGPT responses? Manual sampling of your 20 money queries beats nothing.
• Conversion-weighted traffic. Not raw sessions — sessions that did the thing you wanted. The Overview filters out the curious.
• Share of voice on transactional queries. Overviews are rare down here, the last part of the SERP that behaves the old way.
• Crawl logs, not just Search Console. Are GPTBot, ClaudeBot, PerplexityBot and OAI-SearchBot actually fetching your pages? If they aren't in your logs you're invisible to half the new SERP, and your CDN may be why.

Query type vs SEO strategy for 2027

If you remember one thing, remember which row your target query lives in. A buying-intent page needs comparison and proof, not a 300-word definition at the top.

Where people go wrong (and when to call a pro)

The expensive mistakes are the same unforced errors, repeated across industries.

If you've been running SEO off a 2022 checklist, the gap between that and what the new SERP rewards is wider than it looks. Our services page walks through what we handle.

Frequently asked questions