The clearest AEO pattern from May 18-25, 2026 is that answer engines want more context, not less. Google used I/O week to turn Search into a more persistent conversation. OpenAI added another enterprise data source that ChatGPT can ground answers in. Anthropic bought deeper tooling connectivity. And Google launched a core update right as those interface changes landed, which means source eligibility and ranking stability are now moving at the same time.

That changes the job. The old AEO instinct was to optimize a page for one answer. The current reality is that platforms are building longer sessions, deeper follow-ups, and richer source pools. If your content cannot survive that fuller journey, winning one citation is not enough.

TL;DR

  • Google is pushing Search further into answer-engine behavior. AI Mode now has a new default model, a rebuilt search box, and easier movement from AI Overviews into follow-up conversations.
  • Google's own usage data says AI search is becoming longer, more multimodal, and more planning-heavy. That favors pages built for real tasks, not just isolated keyword matches.
  • ChatGPT and Claude are both moving toward broader data and tool connectivity. For AEO teams, that means public web visibility still matters, but private source pools are becoming part of the answer landscape too.

Is Google turning Search into a full answer engine?

Yes. Google's May 19 product update makes that explicit.

Google is collapsing search, follow-ups, and agent behavior into one surface

Source: Google

Google said it is upgrading AI Mode globally with Gemini 3.5 Flash as the default model and rolling out what it called the biggest Search box upgrade in more than 25 years. The practical detail for AEO is not the branding. It is the flow. Users can move from an AI Overview straight into a back-and-forth AI Mode session, keep their context, and get increasingly relevant supporting links as they go. Google also said the new box can take text, images, files, videos, and Chrome tabs as inputs.

For publishers, that means the old boundary between "ranking for the query" and "supporting the follow-up" is weaker than it was even a month ago. A page now has to hold up across a broader chain of intent. If your article only answers the opener and not the obvious next question, Google is building more ways for the user to keep moving without you.

What to do about it: review your key pages as conversation entry points, not final destinations. Tighten the direct answer near the top, but also make sure the page naturally supports the next comparison, caveat, or implementation question.

Are AI Mode queries getting materially different?

Yes. Google's own numbers say the format of search demand is changing, not just growing.

Google's usage data points to longer, multimodal, decision-stage prompts

Source: Google

Google's May 19 AI Mode usage post gives AEO teams better signal than most product-launch prose usually does. It says more than one in six U.S. searches now use voice or images, that image searches are growing more than 40% month over month, and that the average AI Mode query is triple the length of a traditional Search query. It also says planning-related queries have grown faster than AI Mode overall by 80% over the last six months.

That matters because it shifts what "matching intent" looks like. Answer engines are getting more queries that look like messy project briefs, not neat keywords. The winning page is less likely to be the one that repeats the exact phrase and more likely to be the one that can be cleanly extracted into steps, comparisons, constraints, and tradeoffs.

The immediate move is to test whether your priority pages answer planning-style prompts well. Pages that only work when the user already knows the exact term are now exposed. Pages that handle "help me choose," "what should I do next," and "compare these options" are better aligned with where AI search is heading.

Does the May 2026 core update still matter in an AI-first week?

Yes. It matters more, because ranking volatility now lands inside answer surfaces too.

Google's core systems are still deciding who stays eligible

Source: Search Engine Journal

Google began rolling out the May 2026 core update on May 21, and the only official description so far is that the rollout may take up to two weeks to complete. That sounds routine. It is not routine in context. The update landed two days after Google's big AI Search push, which means traditional ranking movement and AI-surface behavior are now harder to separate in the short term.

For AEO, the point is simple: answer engines do not remove the ranking layer. They add another layer on top of it. If a core update changes which pages Google sees as the most relevant or satisfying, that change can affect both classic clicks and the pages available to AI Overviews and AI Mode follow-ups. You should expect some teams to misread interface changes as citation changes, or citation changes as core-update fallout.

What practitioners should do now is mostly restraint. Do not chase day-one volatility. Keep a clean baseline from before May 21, wait until the rollout settles, and then compare which pages lost rankings, which pages lost citations, and where those two patterns diverge.

Are private source pools becoming part of AEO now?

Yes. Not for every query, but clearly for enterprise answer experiences.

ChatGPT can now ground answers in admin-managed Microsoft Teams data

Source: OpenAI Help Center

OpenAI's May 15 Enterprise and Edu release notes added a Microsoft Teams app with admin-managed sync. Owners and admins can connect Teams once at the workspace level so ChatGPT can reference supported Teams messages and conversation metadata that members already have permission to access. The synced experience is read-only, but that is enough to change the source pool inside enterprise ChatGPT.

This is not consumer web search news, but it is still AEO-relevant. More answer experiences are being grounded in mixed retrieval: public web plus private workspace knowledge. That reduces the share of questions where your published page is the only source competing for the answer. In enterprise contexts especially, the system may now prefer internal memos, support threads, or Teams discussions over your external documentation.

The implication is uncomfortable but useful: external AEO and internal knowledge design are starting to overlap. If your company cares about being cited inside enterprise assistants, governance of internal docs, naming conventions, and message hygiene now matter alongside public SEO.

Why does Anthropic's Stainless deal matter for visibility?

Because better tool connectivity usually means broader retrieval and more grounded answers.

Claude is investing deeper in the systems that connect models to data

Source: Anthropic

Anthropic said on May 18 that it is acquiring Stainless to advance Claude's ability to connect to data and tools, and it tied the move directly to SDKs, CLIs, and MCP server tooling. That is a developer-platform story on the surface. For AEO, it signals something more structural: the answer engines are competing not just on model quality, but on how cleanly they can reach external systems and ground responses in them.

When a platform improves connectivity, it changes what counts as eligible evidence. The strongest answer may come from a help center, a ticketing system, a code repository, a workspace connector, or a structured API response rather than a normal webpage alone. That does not reduce the value of public content. It raises the bar for how well your public and private information ecosystems agree with each other.

What should practitioners do? Treat documentation consistency as a visibility problem. If product pages, help docs, changelogs, and connector-exposed data disagree, you are giving grounded assistants multiple conflicting versions of the same truth.

What gets the easy interpretation wrong?

The easy interpretation is that public-web AEO is being replaced by private connectors and agent systems. We do not think the evidence supports that.

Google's biggest announcements this week still emphasized links to the web, better supporting articles, and more ways to keep browsing through public sources. The core update also underlines that classic search systems are still doing selection work underneath the AI layer. Meanwhile, OpenAI's Teams sync is limited to eligible enterprise and education workspaces, and Anthropic's Stainless deal is a platform signal more than a direct change in consumer answer behavior.

So the better read is narrower. Public-web visibility is still the base layer, but it is no longer the whole layer. Answer engines are getting better at carrying context, handling follow-ups, and pulling from connected systems. If your AEO strategy assumes a single-query, single-page, public-web-only world, it is already behind.

What to do Monday morning

1. Rework your highest-value AEO pages so the first answer is clear and the obvious follow-up questions are answered on the same page. 2. Test key pages against longer planning prompts, not just exact-match queries, and note where the page stops being extractable. 3. Freeze reactive changes during Google's core update rollout and separate ranking shifts from AI-surface changes before making calls. 4. Audit whether your public docs, help content, and any connector-exposed knowledge say the same thing in the same language. 5. If you sell into enterprises, treat internal knowledge cleanliness as part of answer visibility, not just an IT hygiene issue.

What to watch this week

  • Watch whether Google starts surfacing noticeably different supporting links once AI Overview users flow into AI Mode more often.
  • Watch whether brands with strong comparison and decision-stage content benefit more from the shift toward longer planning queries.
  • Watch whether more enterprise answer products add workspace-level connectors, because that will keep shrinking the share of answers driven by the open web alone.

Sources