Key takeaways
- Google's March and May 2026 core updates both caused significant ranking volatility, with the May update completing after 12 days of widespread shifts
- AI Mode and AI Overviews now generate results that can vary by user, meaning traditional rank tracking misses a large portion of your actual visibility picture
- AI Overviews are associated with a 35% reduction in click-through rates for top organic links, making citation presence more important than raw ranking position
- The new Search Console AI reports now show impressions inside AI Overviews and AI Mode -- use them
- Stabilizing visibility in 2026 requires a dual strategy: maintaining traditional SEO health while actively optimizing for AI citation patterns
If you've watched your traffic numbers move in strange directions over the past few months, you're not imagining things. Google has pushed two major core updates in 2026 -- one in March, one in May -- and layered on top of that is the continued expansion of AI Mode and AI Overviews. The result is a search environment where your "rankings" no longer tell the full story of your visibility.
This guide breaks down what's actually happening, why your numbers fluctuate, and what you can do to build a more stable presence across both traditional and AI-generated results.
What happened with the 2026 core updates
Google's May 2026 broad core update launched on May 21 and completed on June 2 -- a 12-day rollout that, by most accounts, hit harder than the March update that preceded it. The volatility wasn't evenly distributed across the rollout period either. Major ranking shifts clustered toward the end of the window, which caught a lot of site owners off guard.
Both updates share a common theme: they reward brands, authoritative sources, and data-rich destinations. Sites that primarily compile or summarize content from elsewhere -- without adding genuine insight or original analysis -- took the biggest hits.
What's worth noting is that Google explicitly confirmed SEO is still the foundation for AI search visibility. That's not a throwaway statement. It means the technical and content fundamentals that have always mattered (crawlability, topical authority, E-E-A-T signals) now feed directly into whether you appear in AI-generated answers, not just blue links.
What the March update established
The March 2026 update set the stage for May by recalibrating how Google's systems evaluate content quality. Sites that recovered from March did so by addressing thin content, improving topical depth, and demonstrating genuine expertise. Sites that ignored those signals got hit again in May.
The pattern is clear: these aren't one-off penalties. They're recalibrations of a system that's increasingly oriented toward surfacing content that AI models can actually use to construct reliable answers.
Why AI Mode makes ranking volatility worse
Traditional rank tracking gives you a position number for a keyword. That number is relatively stable day-to-day (barring updates). AI Mode doesn't work that way.
Google's AI Mode generates personalized responses. The same query from two different users can produce different answers, different cited sources, and different recommendations. This means your visibility in AI Mode isn't a single number -- it's a distribution across user contexts, personas, and query phrasings.
A few specific factors drive this volatility:
Personalization signals. AI Mode incorporates user history, location, and behavior into its responses. A site that ranks well for a general query might not appear in the personalized version of that same query for a specific user segment.
Query fan-outs. When a user asks a broad question, AI Mode doesn't just answer that question. It internally generates sub-queries and pulls from multiple sources to construct a comprehensive response. Your content might be cited for some sub-queries but not others, depending on how well it covers specific angles.
Model updates. Google's underlying AI models get updated independently of core algorithm updates. A change to how the model interprets a topic can shift citation patterns without any corresponding change in your traditional rankings.
Preferred Sources expansion. Google expanded its Preferred Sources feature into AI Overviews in 2026. This means audience loyalty -- users who regularly visit your site -- can now function as a direct visibility signal in AI-generated results. Sites with strong direct traffic and return visitor rates have a structural advantage here.
The click-through rate problem
Here's a number worth sitting with: AI Overviews are associated with a 35% reduction in clicks for top organic links. That figure, cited by marketing analyst Chris Donnelly, reflects a fundamental shift in how users interact with search results.
When Google's AI synthesizes an answer at the top of the page, many users get what they need without clicking through. This doesn't mean your ranking is worthless -- being cited in the AI Overview is itself a form of visibility. But it does mean that ranking #1 in traditional results while being absent from AI-generated answers is a worse position than it used to be.
The new funnel, as Donnelly frames it, runs: Retrieval → Citation → Trust → Conversion. Getting retrieved by the AI model, cited in its response, and trusted enough that users click through to learn more -- that's the sequence that drives traffic in 2026.
How to diagnose your visibility drop
Before you change anything, you need to understand what actually happened. There are three distinct scenarios:
Scenario 1: Traditional rankings dropped
If your organic positions fell during or after a core update window, this is a content quality issue. Pull your Search Console data and look at which pages lost impressions. Ask honestly: does each page deliver genuine value, original insight, or real expertise? Or does it summarize what's already out there?
The path forward is content improvement, not technical fixes. Thin pages need depth. Generic pages need a point of view. Compilation pages need original data or analysis.
Scenario 2: Traditional rankings held but traffic dropped
This is the AI Overview effect. Your position didn't change, but the AI is now answering the query before users reach your link. Check the new Search Console AI reports -- Google now shows impressions inside AI Overviews and AI Mode separately from traditional results.
If you're not appearing in AI Overviews for your target queries, the fix is different from a traditional SEO fix. You need to structure your content so AI models can extract clear, citable answers from it.
Scenario 3: Visibility is inconsistent across queries and time
This is the AI Mode personalization problem. Your content appears for some users and not others, or appears one week and not the next. This is the hardest scenario to diagnose with traditional tools because the variation is real -- it's not a tracking artifact.
To get a handle on this, you need to track your AI visibility across multiple prompt phrasings, user personas, and time periods. Tools like Promptwatch track how AI search engines actually behave in real user interfaces -- not just through API calls -- which matters because the answers users see can differ meaningfully from what API queries return.
What actually stabilizes AI Mode visibility
Stabilizing your presence in AI-generated results isn't about gaming the system. It's about making your content genuinely useful to the AI models that construct those answers. Here's what that looks like in practice.
Build topical authority, not just keyword coverage
AI models favor sources that demonstrate deep, consistent expertise on a topic. A site with 50 thorough articles on a specific domain will be cited more reliably than a site with 500 shallow articles covering everything.
Map your content against the questions AI models are actually answering in your space. Where are the gaps? Which angles are competitors covering that you're not? Answer Gap Analysis -- looking at which prompts your competitors appear for but you don't -- is one of the most direct ways to find these opportunities.
Structure content for extraction
AI models don't read your content the way humans do. They extract specific claims, definitions, and answers. Content that's structured to support this extraction gets cited more often.
Practically, this means:
- Use clear, direct answers near the top of sections (don't bury the lede)
- Include definitions for key terms
- Use structured data where appropriate (FAQ schema, HowTo schema)
- Write in a way that makes individual sentences quotable and accurate
Fix technical crawlability for AI agents
Traditional SEO crawlability (Googlebot can index your pages) is necessary but not sufficient. AI citation agents -- the systems that pull content to construct AI responses -- have their own crawl behavior. They may hit your pages more frequently, encounter JavaScript rendering issues, or fail to access content behind certain authentication patterns.
Monitoring which pages AI crawlers are actually reading, what errors they encounter, and how often they return gives you a layer of insight that traditional SEO tools don't provide.
Diversify your citation surface
If your AI visibility depends entirely on your own website, you're exposed. AI models pull from a wide range of sources: Reddit discussions, YouTube videos, third-party listicles, review sites, and industry publications. Being present and accurate across these sources makes your brand more likely to appear in AI responses, even when the AI doesn't cite your site directly.
This is sometimes called "offsite citation optimization" -- making sure that the external web's representation of your brand is accurate, positive, and present in the places AI models draw from.
Monitor AI visibility separately from traditional rankings
This is probably the most important operational change. If you're only tracking traditional keyword rankings, you have a significant blind spot. AI Mode visibility requires different tracking: which prompts trigger citations of your content, which AI models cite you, how often, and for which topics.
Promptwatch covers 10 AI models including Google AI Mode, AI Overviews, ChatGPT, Perplexity, and others. The page-level tracking shows exactly which of your pages are being cited and by which models -- which makes it much easier to understand what's working and what isn't.
Tools worth using in 2026
Here's a quick comparison of tools relevant to diagnosing and improving AI Mode visibility:
| Tool | Traditional SEO | AI visibility tracking | Content optimization | Crawler monitoring |
|---|---|---|---|---|
| Google Search Console | Yes | Partial (new AI reports) | No | No |
| Screaming Frog | Yes | No | No | Partial |
| SE Ranking | Yes | Partial | Yes | No |
| Clearscope | No | No | Yes | No |
| Promptwatch | No | Yes (10 models) | Yes (Content Agents) | Yes |
| AccuRanker | Yes | No | No | No |
For traditional SEO health, Google Search Console combined with a crawler like Screaming Frog covers the basics. For AI visibility specifically, you need a dedicated tool -- the data just isn't available through traditional SEO platforms.
The Preferred Sources signal: what it means for your strategy
Google's expansion of Preferred Sources into AI Overviews in 2026 introduced something genuinely new: audience loyalty as a ranking signal for AI-generated results. If users regularly visit your site directly, that behavior can influence whether your content appears in AI Overviews for those users.
This has a few practical implications:
Your email list and direct traffic matter more than they used to. Users who come back to your site repeatedly are building a signal that Google can use to personalize AI results in your favor.
Brand search volume matters. When people search for your brand name, that's a signal of authority and trust. Building brand awareness -- through PR, social, community, or any other channel -- now has a more direct connection to AI search visibility than it did two years ago.
Content that generates return visits (tools, calculators, reference pages, regularly updated resources) builds this signal more effectively than one-and-done articles.
What to do right now
If you're dealing with a visibility drop and want to stabilize, here's a practical sequence:
-
Check Search Console's new AI reports to separate traditional ranking drops from AI Overview drops. They require different responses.
-
Audit your highest-traffic pages that declined. For each one, ask whether it provides genuine value that couldn't be found in 30 seconds elsewhere.
-
Identify the prompts and questions your target audience is asking AI models. Map those against your existing content. The gaps are your content priorities.
-
Fix any technical issues that might prevent AI crawlers from reading your content -- JavaScript rendering problems, slow load times, or pages blocked by robots.txt.
-
Start tracking AI visibility with a dedicated tool so you have a baseline. You can't improve what you can't measure, and traditional rank trackers won't show you what's happening in AI Mode.
The core updates of 2026 didn't change what good content looks like. They changed how much it matters that your content is genuinely good -- and they added a new layer of visibility (AI Mode) where that quality gets tested in a completely different way. Sites that treat these as separate problems will keep chasing their tails. Sites that build for both will find the volatility becomes much more manageable.