Google's own Search Central documentation states that AI Mode and AI Overviews "may use different models and techniques, so the set of responses and links they show will vary." They are not the same feature at two sizes. They have different triggers, different intents and, critically, different measurement in Search Console. AI Mode reports impressions only, with no clicks and no CTR, which means a large part of your Google AI performance is currently unmeasurable by design.
Key Takeaways
- Google documents these as different systems, not variations of one. The link sets diverge because the models and techniques diverge.
- AI Overviews trigger when Google's systems judge that a generated summary adds value, and are framed as a jumping-off point to links.
- AI Mode is for further exploration, reasoning and complex comparisons, a different user intent entirely.
- Both use query fan-out, which is why both surface a wider and more diverse set of links than the classic ten blue links.
- Controls are
nosnippet,data-nosnippet,max-snippetandnoindex. Google-Extended governs training and grounding in other Google systems, not Search AI features. There is no way to opt out of AI Overviews while staying in Search. - Search Console reports impressions only for AI Mode, no clicks, no CTR, and a logging error between 13–17 August 2026 corrupted AI impression data. An August dip may be an artefact.
Start With What Google Actually Says
There is enough speculation about these features that it is worth anchoring on the primary source. Google's guidance on AI features in Search lives in Google Search Central, and the load-bearing sentence is this: AI Mode and AI Overviews "may use different models and techniques, so the set of responses and links they show will vary."
That single sentence invalidates a common workflow: checking your AI Overview presence for a query and assuming it tells you your AI Mode presence. It does not. They are separate surfaces with separate outcomes, and you have to check both.
The Comparison, Laid Out
| AI Overviews | AI Mode | |
|---|---|---|
| What it is | A generated summary at the top of a standard results page | A dedicated conversational surface for exploration |
| When it appears | When Google's systems judge a summary adds value to the query | When the user chooses it for further exploration |
| User intent served | Quick orientation on a question | Reasoning, comparison, multi-step exploration |
| Framing in Google's docs | A jumping-off point to links | Deeper exploration and complex comparisons |
| Retrieval approach | Query fan-out | Query fan-out |
| Link diversity | Wider than classic results | Wider than classic results |
| Models and techniques | May differ from AI Mode | May differ from AI Overviews |
| Search Console: impressions | Reported | Reported |
| Search Console: clicks | Reported | Not reported |
| Search Console: CTR | Available | Not available |
| Can you opt out and stay in Search? | No | No |
| Snippet controls apply? | Yes | Yes |
| Governed by Google-Extended? | No | No |
The bottom four rows are where most of the practical consequences live.
Query Fan-Out Is the Mechanism Behind Both
Both features expand a single user query into multiple related sub-queries, retrieve for each, and synthesise across the results. That is query fan-out, and it explains a phenomenon a lot of teams find confusing.
Why you appear for queries you never targeted
If someone asks a broad question, fan-out might generate a dozen narrower sub-queries. Your page ranks for one of those sub-queries, a specific comparison, a definitional question, a niche edge case, and gets pulled into an answer for a query you have never ranked for and never optimised toward.
This is generally good news. It means depth on specific sub-topics earns you inclusion in broader answers.
Why the link set is more diverse
Fan-out retrieves across multiple sub-queries, so the resulting citation set draws from more distinct sources than a single ranked list would. Practically, more sites get a slice, but each slice is smaller and less predictable.
Why this rewards coverage over singular pages
The strategic implication is that a cluster of genuinely useful pages covering the sub-questions in your topic will outperform one heavily optimised flagship page. Fan-out is a breadth mechanism, and it rewards breadth.
What You Can and Cannot Control
This is where I see the most confident wrong advice, so let us be precise.
The controls that work
`nosnippet` prevents a text snippet from being shown for the page. `data-nosnippet` does the same at the element level, so you can exclude a specific block while leaving the rest of the page eligible. `max-snippet` caps the snippet length. `noindex` removes the page from Search entirely, which necessarily removes it from AI features too.
These limit what can be shown. They do not give you a switch for AI features specifically.
The control that does not do what people think
Google-Extended governs training and grounding in other Google systems, not Search AI features. Adding Google-Extended to robots.txt does not remove you from AI Overviews or AI Mode.
I have seen this exact mistake in production on sites whose teams believed they had opted out. They had not. They had made a training-data decision and left their Search AI presence entirely intact.
The thing you genuinely cannot do
There is no mechanism to remain in Google Search while opting out of AI Overviews. The only route to exclusion is noindex, which removes you from Search altogether. That is the trade, stated plainly, and no configuration trick changes it.
Anyone selling you an "AI Overviews opt-out" that keeps your organic rankings is selling something that does not exist.
The Measurement Gap Is the Real Story
Here is the part that should change how you report internally.
Search Console gives you impressions for AI Mode. No clicks. No CTR.
What that actually means for your reporting
You can see that you were shown. You cannot see whether anyone acted on it. Any performance narrative you build about AI Mode is therefore built on half the data, and you should say so out loud when presenting it.
I would rather tell a founder "we have impression data and no click data for this surface, so here is what we can and cannot conclude" than produce a confident chart that quietly assumes the missing half.
The 13–17 August 2026 logging error
Between 13 and 17 August 2026, a Search Console logging error corrupted AI impression data. If you are looking at an August 2026 dip in AI impressions and building a story about it, check your dates first. You may be analysing an artefact.
This is the kind of thing that gets a strategy changed on the basis of a bug. Annotate the window in your reporting so nobody comes back to it in six months and draws the wrong conclusion.
How to build reporting that survives this
Three practical adjustments. Annotate known data-quality windows directly in your dashboards. Report AI Mode impressions as a visibility indicator rather than a performance metric. And triangulate with what you can measure: total organic sessions, branded search volume, direct traffic, assisted conversions, rather than leaning on a single incomplete number.
Recent Developments Worth Tracking
AI Mode has been moving quickly. As of August 2026, one development is worth flagging: link carousels for trending topics, where Preferred Sources can appear.
Preferred Sources is Google's mechanism for users to indicate publishers they want to see more of. If Preferred Sources selections influence carousel placement on trending topics, then the value of a reader deliberately choosing your site goes up: and that is earned through consistent quality and brand affinity, not through technical configuration.
For fast-moving reporting on these changes, Search Engine Journal and Search Engine Land are the places I check. Google's own release notes remain the authority when the two disagree.
What I Would Do Differently for Each Surface
Same underlying site, different emphasis.
For AI Overviews
Optimise for extractable, clearly stated answers. AI Overviews are framed as a jumping-off point, so the goal is to be the source worth clicking through to. Direct answers, clean structure, a reason to go deeper.
For AI Mode
Optimise for depth and comparison. AI Mode serves reasoning and complex comparisons, which means content that handles nuance, trade-offs and edge cases has an advantage. Comparison tables, honest limitations, decision frameworks.
For both
Cover the sub-questions. Fan-out is the shared mechanism, and a topic covered in depth across several genuinely useful pages gets pulled into more answers than a single page can.
Frequently Asked Questions
Are AI Mode and AI Overviews the same feature?
No. Google documents them as potentially using different models and techniques, with the result that the responses and links they show will vary.
Why do I appear in one but not the other?
Because they are separate systems with separate retrieval. Presence in one tells you nothing reliable about presence in the other.
Can I opt out of AI Overviews and stay in Google Search?
No. The only exclusion mechanism is noindex, which removes you from Search entirely.
Does blocking Google-Extended remove me from AI Overviews?
No. Google-Extended governs training and grounding in other Google systems, not Search AI features. This is a widespread and costly misunderstanding.
What controls actually limit what appears?
nosnippet, data-nosnippet, max-snippet and noindex. They constrain what can be displayed; they are not an AI-feature toggle.
Why does Search Console not show AI Mode clicks?
It currently reports impressions only for AI Mode: no clicks and no CTR. That is the present state of the reporting, not a configuration issue on your end.
My AI impressions dropped in August 2026. Is that real?
Check the dates. A Search Console logging error between 13 and 17 August 2026 corrupted AI impression data. A dip covering that window may be an artefact.
What is query fan-out?
The expansion of one user query into multiple related sub-queries, each retrieved for separately, with results synthesised into one answer. It is why AI features surface a wider and more diverse set of links.
Does ranking first still guarantee citation?
No. Fan-out means the citation set is assembled across many sub-queries, so a page ranking well for a narrow sub-question can be cited in an answer where the classic top result is not.
What are Preferred Sources and do they matter?
Preferred Sources lets users indicate publishers they want to see more of. As of August 2026, Preferred Sources can appear in AI Mode's trending-topic link carousels, which raises the strategic value of readers actively choosing your site.
If any of this is landing awkwardly against your own reporting: the impressions-without-clicks problem in particular tends to create difficult conversations with leadership, I am happy to talk it through. I have spent 4+ years in marketing helping edtech and startup brands grow organically, including the work behind Masai School's Instagram going from 26K to 117K and LinkedIn from 50K to 160K. You can browse more of that work and reach me via the contact form at younusfardeen.com, whenever it is useful.