Google's May 2026 Core Update and AI Search

Google's May 2026 core update finished rolling out on June 2, after twelve days of heavy ranking volatility that started on May 21. It was the second core update of 2026, and it landed in the middle of the biggest AI search overhaul Google has shipped. The part most teams missed: this one moved AI citations, not only blue links. Pages that held their ten-blue-link position still disappeared from AI Overviews, and pages sitting on page two started showing up as cited sources inside AI answers.

A Google core update is a broad change to Google's core ranking systems that re-scores how trustworthy and relevant your content looks across the entire index. In 2026 that re-scoring feeds two surfaces at once: the traditional results and the AI answers (AI Overviews, AI Mode, and the assistants that pull from Google's index). For a B2B SaaS team, recovery now means winning on both, which is the work we call AEO (Answer Engine Optimization).

This is the analysis we ran for ourselves and our clients while the update was still settling: what changed, what our own Search Console data showed, who got hit, and the exact playbook to recover.

On this page

• What was the Google May 2026 core update?
• Why did a core update move AI citations and not only rankings?
• What did our own data show during the rollout?
• Who got hit hardest by the May 2026 core update?
• How do you recover? The AEO playbook
• How long does core update recovery take?
• Common questions

What was the Google May 2026 core update?

Google confirmed the May 2026 core update on May 21, 2026, and marked it complete on June 2, roughly twelve days later. Google's standard line held: a regular update designed to better surface relevant, satisfying content from all types of sites. The reality on the ground was sharper. Trackers logged major volatility the weekend of May 23, then another spike around June 2 as the rollout closed, reported by Search Engine Land and Search Engine Roundtable.

Three facts matter for planning:

• This was the second core update of 2026, arriving weeks after a March update that was one of the most volatile on record.
• It shipped alongside Google's largest expansion of AI Overviews and AI Mode, so ranking changes and AI-answer changes happened together.
• Core updates are sitewide quality re-scoring, not page-level penalties. There is no single 'fix this tag' lever. Recovery is a content and structure problem.

Key takeaway: the May 2026 update was a broad, high-volatility re-scoring that hit traditional rankings and AI answers in the same window.

Why did a core update move AI citations and not only rankings?

For years, ranking in the top ten was a reliable proxy for getting cited in an AI answer. That link is breaking. Ahrefs studied 863,000 SERPs and 4 million AI Overview URLs and found that only 38% of AI Overview citations now come from pages ranking in the top ten organic results, down from 76% in their July 2025 study (Ahrefs, covered by Search Engine Journal). The rest split almost evenly between positions 11 to 100 and pages outside the top 100.

That gap is the whole story. AI answers are a separate selection layer sitting on top of the index, and they choose sources on different signals than the ten blue links do. When Google re-scores the index in a core update, it shifts both layers, and the AI layer can move further than the ranking layer.

What the AI layer rewards comes down to four things, the same mechanics we use to evaluate any AEO program:

1. Being in the index. AI crawlers have to reach the content and your architecture has to expose it. If a page is buried or blocked, it cannot be cited.
2. Being extractable. AI answers favor passages that answer the question directly and are easy to lift: a clear answer up top, question-shaped headers, short declarative sentences, tables, and structured data that labels what each thing is. We went deep on this in our piece on what makes a page extractable for RAG-based answer engines.
3. Being trusted. Models lean on sources they already weight, which is partly your own authority and partly where else you are mentioned. Verifiable claims, real authors, and consistent entity data win.
4. Being measured. Tracking share of voice and citations inside the engines, not only Google positions.

The May 2026 update tightened the screws on points two and three. Industry analyses of AI Overview source selection through the update describe engines getting more selective, favoring pages with verifiable claims and clean structured data over longer pages with weaker evidence. Citations also moved inline, attached to the specific sentence or bullet the answer draws from, so a single AI Overview can cite three different sources for three different claims.

Key takeaway: a high ranking no longer guarantees an AI citation, so a core update can quietly erase your AI visibility even when your blue-link position survives.

What did our own data show during the rollout?

We watched it happen on our own site in Search Console. Two honest caveats first: karpi.studio is a small site, so the daily numbers are small, and our clean before-and-after snapshots cover the opening days through the May 23 volatility spike, not the fully settled picture. With that said, the pattern was clear enough to act on.

Through the start of the rollout, our site behaved like most sites did. Average position softened from the 8 to 10 range into the low teens on May 23, and clicks dipped on the worst volatility day. We felt the update.

The interesting part was which pages held. Our Schema Glossary entries, the pages built specifically to be machine-readable and entity-clear, kept pulling impressions and converted clicks at 10 to 25 percent click-through from positions in the 4 to 12 range, while broader pages wobbled. Small numbers, one site, the opening days of a twelve-day rollout. Directionally it matched the macro signal exactly: when ranking position moved, the pages engineered for extraction held their ground.

That is the entire argument for treating structure as a ranking and citation asset, not a nice-to-have. We are not claiming structured data is a magic citation button. We are saying the pages we built to be clean, labeled, and answer-first are the ones that survived the shake.

Key takeaway: across our own pages, structure and extractability held up better than raw ranking position during the update.

Who got hit hardest by the May 2026 core update?

The losers were consistent with every recent core update, sharpened by the AI layer:

• YMYL sites took the worst of it. YMYL stands for Your Money or Your Life: content that can affect a person's health, finances, safety, or major life decisions. Google holds it to the highest E-E-A-T standard because wrong information there can cause real harm, so health, finance, legal, and home services pages get re-scored hardest. Thin or generic content in those categories dropped sharply. If you are in healthtech, this is not optional. The conversion and trust work we did for Blueberry Pediatrics, a telehealth company, lives in this YMYL zone.
• Thin and AI-generated filler lost ground. High volumes of generic content with no first-hand experience faced scrutiny, in and out of YMYL niches.
• Pages with weak authorship. Google's E-E-A-T framework now treats Experience as the most important pillar. Content from a named author with verifiable credentials and real first-hand experience outperformed anonymous or ghost-written pages.

The winners were sites with genuine topical depth, coherent information architecture, and content that adds new knowledge rather than restating what already ranks (a signal often called Information Gain).

Key takeaway: YMYL, thin content, and anonymous pages lost the most, while deep, experienced, well-structured content gained.

How do you recover? The AEO playbook

Core update recovery is not a checklist of tags. It is raising the quality and machine-readability of the site until both Google and the AI layer trust it again. Here is the sequence we run, mapped to the four mechanics above. This is the same work behind our answer engine optimization service.

1. Fix the index layer first. Confirm AI crawlers can reach your content, flatten any architecture that buries key pages, and ship an llms.txt file. If the engines cannot read it, nothing else matters.
2. Make every important page extractable. Lead with a direct answer, use question-format headers, keep sentences short and declarative, add tables for comparisons, and deploy clean structured data (FAQ, How-To, Product, Article, author). Our guide on why schema is the foundation of AEO and the deeper schema and CMS structure walkthrough cover the how.
3. Prove experience and expertise. Put a real author on every page with credentials and a bio, write from first-hand experience, and add genuinely new data, examples, or analysis instead of summarizing the current top results. This is the E-E-A-T work the update rewarded most.
4. Back every claim with evidence. AI answers now favor verifiable claims. Add specific numbers, name your sources, link them inline, and cut the vague assertions. Pages with weak evidence lost citations to pages with strong evidence.
5. Build off-site trust. Earn mentions on the sources models already weight in your category (industry publications, Reddit, YouTube, G2, real PR). Consistent entity information across the web makes you easier to trust and to tell apart from similarly named companies.
6. Tighten accessibility. Semantic HTML, proper heading order, descriptive alt text, and clean markup help screen readers and AI parsers read the same way. Accessibility and machine-readability are the same muscle.
7. Measure inside the engines. Track AI share of voice and citations alongside Google position, so you can see recovery on the surface that moved most.

We have run this both ways. The same mechanics grew dYdX organic search by 3,722 percent during a Webflow migration. Closer to the point of this article, we took karpi.studio from zero to about 150 AI citations a day in three months by building exactly the structure and authority described above. Same mechanics, both times.

Key takeaway: recover by fixing crawlability, extractability, experience signals, evidence, off-site trust, accessibility, and AI-layer measurement, in that order.

How long does core update recovery take?

Recovery is slow and not guaranteed. Sites that make real content and structure improvements can see partial gains within 6 to 8 weeks. Full recovery typically takes 3 to 6 months and often does not complete until the next broad core update re-scores the index. Google has said plainly that not every site fully recovers. The teams that come back are the ones that treat the update as a quality reset rather than a tag to patch.

If your AI Overview citations dropped while your rankings held, or your rankings dropped and you have no idea which pages still get cited in AI answers, that is the gap we close. See our answer engine optimization service and our roundup of the Webflow AEO agencies worth knowing if you want to compare options before you talk to anyone.

Common questions

Did the May 2026 core update affect AI Overviews and ChatGPT?

Yes. The update re-scored Google's index, which feeds AI Overviews and AI Mode directly, and feeds assistants that pull from Google. Because AI answers select sources on extractability and verifiable authority rather than ranking alone, many sites saw their AI citations move more than their blue-link positions. Ahrefs found only 38 percent of AI Overview citations now come from top-ten pages, down from 76 percent in 2025.

How do I know if the core update hit my site?

Compare the four weeks before May 21 to the period after June 2 in Google Search Console, by page and by query. Look for clicks and position drops on your money pages. Then check the AI layer separately: search your key questions in AI Overviews, ChatGPT, and Perplexity and see whether you are still cited. The two can move in different directions.

Is recovering from a core update the same as doing AEO?

Largely, yes, in 2026. A core update re-scores quality, trust, and relevance. AEO is the practice of structuring content so both Google and AI engines can trust and extract it. The recovery work (extractability, E-E-A-T, evidence, structured data, off-site trust) is the same work that earns AI citations in the first place.

What kind of pages recovered fastest?

Pages with a clear answer up top, real authorship, specific evidence, and clean structured data. On our own site, the entity-clear Schema Glossary pages held position and click-through through the volatility while broader pages dipped. Depth and structure beat thin content built for keywords.

Methodology and disclosure

This analysis combines public reporting on the update (Search Engine Land, Search Engine Roundtable, Search Engine Journal), the Ahrefs AI Overview citation studies (863,000 SERPs and 4 million AI Overview URLs), and our own Google Search Console data across the rollout window. Our first-party data comes from a small site and covers the opening volatility spike, not the fully settled result, so we have flagged it as directional. Every external figure is attributed to its source and linked inline. This reflects our read as of June 2026. If something here is wrong, email pavel@karpi.studio and we will correct it.

By Pavel Karpisek. Founder of Karpi Studio, Premium Webflow Enterprise Partner with 6+ years on the platform and 200+ B2B projects shipped. Last updated: June 2026.