What to do now that AIOs turned search into reading sessions
Intent compression breaks the segmentation SEO ran on for 20 years. Your content strategy map survives, and your per-page optimization might not.
Intent still tells you what to write. But when an AIO lands on the SERP, users no longer behave the same way as classic search.
In this memo:
AIO compresses 5 distinct search intents into one reading pattern
What winning the “second impression” looks like for product, category, and blog pages
The 1-slide explanation that reassures stakeholders the content team’s last 3 years were not wasted (premium)
A Claude skill that audits your meta descriptions against the competitors sharing your SERP (premium)
Scale up your brand visibility, everywhere search happens
Semrush for Enterprise empowers businesses to own every layer of brand visibility.
By unifying SEO, AI search, and technical site health into one platform, teams gain comprehensive performance tracking and complete optimization capabilities. Swap busywork for impactful strategy with advanced automations that are backed by the market’s leading search database.
It’s how winning brands ensure they show up wherever their customers search.
Explore Semrush for Enterprise
The new mental model of search intent
Last week, I shared how Eric Van Buskirk of Clickstream Solutions and I analyzed anonymized clickstream data from approximately 846K U.S.-based Google search sessions.
The most significant finding? The time-on-page for a user on the SERP is no longer dependent on search intent when an AIO is present. The AIO compresses search intent behavior to look similar across intent types.
Old mental model of search intent: Navigational searches are “fast;” informational ones are “slow.” Time-on-SERP follows intent, and SERPs without an AIO clearly show this pattern (similar to classic search pre-AI outputs), demonstrated by 12% of navigational searches vs. 32% of local searchers still being on the SERP after 21 seconds.
New mental model of search intent: There is barely a difference in how long users spend time on SERPs between user intents when there is an AIO present. 42-48.5% of users are still on the SERP after 21 seconds across all the 5 major intents.
From The same user behaves differently in AIOs vs. AI Mode (bold added):
At 21 seconds into a session without an AIO, only 12% of navigational searchers are still on the page. 32% of local searchers are. In classic search, time-on-page has always followed intent: navigational users leave fast because they know where they’re going, local users stay because the SERP is dense with maps and listings, informational users fall somewhere in between.
With an AIO present, the spread compresses to barely 6 points. All five intent types (informational, local, navigational, transactional, video) cluster between 41.9% and 48.5% time-on-page at 21 seconds.
Notice how much longer average SERP sessions are - almost 4x! So, we can conclude that AI Overviews don’t just compress user intent but also prolong the time users spend with search results.
The reason? Additional context. Direct answers from the AIOs provide more information and take longer to read. The intent behind the initial query matters less.
This is the gap between links and answers in the new AIO-filled SERP. In the past, giving users a list of (ten blue) links meant the user was responsible for verifying accuracy and finding the information after the click through. Therefore, Google gets user feedback from their behavior.
But when Google (or other LLMs) gives the answer directly, that onus is on the answer engine.
Bing’s blog Evolving the role of the index brings this to a point:
Grounding an AI–generated answer introduces a fundamentally different constraint: The system is no longer just pointing to information, it is using it. The goal shifts from “fetch the best documents” to “fetch the best information to synthesize into a reliable, verifiable answer.”
Lastly, it also means there is utility in tracking branded prompts more diligently and making sure LLMs return the desired information about a brand. Just like companies bid on their brand as a defense mechanism, they should monitor branded prompts, not just product- or painpoint-related ones.
Heads up: The Adapting for AI-Mode Based Search deck gives directors a validated, data-backed story to present to executives who keep asking what changed. Find it in the Premium Subscribers Resource Library.
Why this matters
For 20 years, what you searched told Google and SEOs how you’d behave. Type a brand name (navigational search), and you’re in and out in seconds. Search “best CRM for startups” (comparison search), and you settle into a set of comparison pages. Intent sorted everyone.
The AIO erased that tell. By dropping a block of answer text at the top of the page, it pulls every searcher into a reading session, no matter why they came. The brand-name searcher reads the AIO. The product researcher reads another. Both slow down, both stay, both behave alike on the SERP page. That flattening is the intent compression.
Most Google users never chose this, because most Google users are not AI early adopters. They meet AI through Google’s search results as Google forces guides them into AIOs and AI Mode at the top of the results. AI is changing its search behavior passively and without searchers’ explicit consent. Many Google users might not even realize they’re using AI, and that’s one reason we’re seeing growing installs of DuckDuckGo.
Google reports more than 1.5 billion people use AI Overviews, so this is not an edge case. It’s how the web is searching now.
More than 1.5 billion users around the world use AI Overviews for help with their questions. Source: Google
What winning the second impression looks like
The AI Overviews vs. AI Mode behavior analysis revealed the significance and optimization opportunity of the second look. The second impression is what searchers see on the back-scroll, after they’ve already passed your listing once. It’s like a double-take made by a grocery shopper in the cereal aisle who has dozens of options: The shopper scans every box in view, then circles back to reread the one that caught their eye.
Metadata is the trigger for selecting search results. Rich snippets catch attention early on, but they might not be enough to convert users to a click, especially if new search behavior has shifted to include a thorough reading of the SERP and a second scroll up. What can earn the click is what shows up next to your listing on that second pass, and it has to be relevant and trustworthy.
Different page types require different relevance and trust indicators.
Product detail pages (PDP)
Users compare star ratings, review count, price, and stock status.
3 things to control outside the meta:
Product schema with aggregateRating, review, offers, and availability. Miss any one of these, and a competitor’s listing renders fuller than yours.
Review count is a comparison field. 47 reviews next to a competitor’s 2,300 loses on the second pass even if your description is sharper. Review velocity is a competitive moat.
Multiple images in the schema array so Google has options for different SERP layouts.
Category detail pages (CDP)
Category pages compete with the AIO’s own list. If the AIO already enumerated 5 options, your category page has to look like the place where the user goes to actually choose between them.
3 things to control outside the meta:
ItemList schema on the category page so Google can render product carousels in the SERP. A carousel takes more vertical space than a single listing and dominates the back-scroll.
Filter and sort UI visible in the SERP preview. Google sometimes surfaces sitelinks for category facets (”by price” or “by brand”). Internal linking to those facets makes them eligible.
Page count and depth. A category page with 12 products competes badly against one with 240. The second impression carries an implicit “Is this comprehensive?” check.
Blog content
The AIO has already given the user the answer. To earn a possible “validation click”, or at least a thorough second impression for your brand, what the user is looking for is credibility on who said it and when.
2 things to control outside the meta:
Visible datePublished or dateModified in the SERP. A 2024 article next to a 2026 article likely loses, whether the user considers the description or not.
Article schema with a named author field that links to a sameAs URL (LinkedIn, author bio page). This makes the author an entity Google can resolve, which matters for E-E-A-T scoring even if no visible card renders.
The AI SEO Change Management Plan alone saves directors 10+ planning hours for retraining your team against current best practices, but Premium is only $150/year. See the full library.
What intent compression means for operators
The last 3 years of intent-based content work produced the right pages for the right queries. What’s shifting is one prediction layer on top of that strategy: how long users stay, where they look, when they click.
That layer is now AIO-driven, not intent-driven. User search intent still drives what your brand needs to write, but it is no longer a good estimate of how users will behave on a SERP page.
Therefore, more optimization work moves to the SERP, focusing on how your listing reads against the AIO and the results around it, separate from the intent logic that decides what the page content should be to answer the query.
Your core optimization efforts don’t change. A content team built around intent clusters keeps its cluster map intact. What gets updated is the optimization pattern per page (meta descriptions, title tags, the second-impression framing from finding 7 shared last week with Premium subscribers), not the underlying taxonomy or content strategy.
(Premium-only) Claude skill for identifying 3+ competitor words at scale
Last week, I instructed Premium Growth Memo readers on how to optimize for more competitive metadata within the SERP.
Today, I’ve included a Claude skill that’s ready to do that for you.
Also included: A short helpful explainer to educate your stakeholders on the user-behavior changes we found across the AIO-based SERP.
