Google SGE Is Now AI Mode: What It Means for Traffic

• Google SGE was retired as a name in 2026 and split into two surfaces: AI Overviews, a synthesized answer above the results, and AI Mode, a separate conversational search tab.
• AI Overviews now appear on roughly half of US searches and cut organic click-through by 34% to 61% when present, so the loss falls hardest on informational queries.
• A December 2025 Surfer SEO analysis of 173,902 URLs found about 68% of pages cited in AI answers do not rank in the top 10 for the query, so citation and ranking are now separate outcomes.
• AI Mode builds answers with query fan-out: it breaks one question into many sub-queries, so pages that cover a whole topic cleanly get pulled in more often than pages built around a single keyword.

Google SGE was the label for Google's first attempt at answering searches with generative AI, and in 2026 it quietly became two products: AI Overviews and AI Mode. The name retired, the behavior scaled. AI Overviews now show up on close to half of US searches, and independent analysis puts the click-through drop, when one appears, at as much as 61%. That is the number most operators already feel in their analytics.

Here is what the headline misses. Losing clicks on informational queries is not the same as losing revenue, and the pages that get quoted inside these AI answers are often not the pages ranking in the top ten. Google changed both how it reads the web and how it shows results, so ranking first and getting traffic have come apart. This guide covers what SGE became, how much traffic AI Overviews and AI Mode actually redirect, how the system chooses its sources, and what to change in your SEO so you stay visible where the answer gets written.

What Google SGE Became: AI Overviews and AI Mode

The short version: Search Generative Experience, or SGE, was the 2023 test. Google graduated it, dropped the SGE name, and shipped the capability as two separate features that do different jobs.

AI Overviews change what you see: a short AI answer stitched above the blue links. AI Mode changes how you search: a dedicated tab where one question turns into a back-and-forth conversation with follow-ups.

The distinction matters because each surface rewards content differently. One favors a clean, extractable answer near the top of a page. The other rewards topical completeness, because it asks several questions at once and assembles the reply from many sources. If you are still weighing where AI search sits next to classic rankings, our breakdown of how GEO and SEO differ sets the frame this article builds on.

Dimension	AI Overviews	AI Mode
What it is	AI summary above the standard results	A separate conversational search experience
Where it lives	Inside the normal results page	Its own tab, opened on purpose
Query type	Straightforward, single questions	Complex, multi-part research with follow-ups
How it answers	One synthesized block with a few citations	Many parallel searches merged into one reply
What content wins	Clear answer blocks it can lift	Pages that cover the full topic, not one keyword

Both pull from the live index rather than answering from memory, and both cite a small set of sources. Your goal shifts from owning a position to being one of the pages the model reaches for.

How Much Traffic AI Overviews Actually Take

The click erosion is real, but uneven. AI Overviews do not touch every query the same way, and the sites that panic tend to be the ones that lived on informational traffic that was never going to convert. Here is what the current data shows.

• Coverage keeps climbing: AI Overviews surface on roughly 47% to 64% of searches depending on the study, and near 50% in the US.
• Click-through falls 34% to 61% on queries where an AI answer appears, with positions 1 to 3 taking the sharpest hit.
• Health, finance, and how-to topics trigger AI answers most often, while local and transactional searches see far lower coverage.
• The clicks that do come through convert better, because the reader who still clicks has a stronger, more specific intent.

Ranking first no longer guarantees the click. When AI answers the question in place, the value moves to being the cited source inside the answer, not the top link beneath it.

Where the traffic loss lands hardest

Definitional and quick-fact queries are the most exposed, since the AI answer fully satisfies the searcher. Commercial and comparison queries hold up better, because people still want to see options, prices, and proof before they act. Read your own analytics by query type before assuming the drop is uniform, and separate the zero-click searches you lost from the revenue queries you still own.

How AI Mode Picks Its Sources: Query Fan-Out

AI Mode does not run your query once. It uses a retrieval method Google calls query fan-out, which turns a single prompt into a set of related sub-queries and runs them in parallel. Understanding the flow tells you exactly what to write for.

1. Decompose the prompt. The question is parsed for entities, constraints, and intent, then split into several narrower sub-queries.
2. Retrieve in parallel. Each sub-query hits the live index, the Knowledge Graph, and structured data at the same time, widening the pool of candidate pages.
3. Expand across sources. The system gathers passages from many pages rather than one, favoring content that answers a specific slice cleanly.
4. Synthesize and cite. It merges the passages into one answer and links the handful of pages it actually used.

One visible question can trigger a dozen hidden ones. A page that answers the whole cluster of related sub-questions gets cited across more of those fan-out queries than a page aimed at one exact-match term.

This is why thin, single-keyword pages disappear from AI answers while thorough, well-structured pages keep showing up. The practical move is to write pages AI wants to quote: self-contained answer blocks, clear definitions, real numbers, and headings that mirror the questions people actually ask.

What Changes for Your SEO Playbook

The foundation does not change, the target does. Technical health, crawlability, and genuine authority still decide whether you are even in the running. What changes is how you shape content and what you measure. Focus the work here.

• Answer first, then expand. Open each section with a direct, quotable answer of 30 to 50 words, then add the context. That block is what gets lifted.
• Cover the cluster, not the keyword. Map the sub-questions around a topic and answer them on one page, so fan-out retrieval keeps finding you.
• Lead with data. Specific numbers, sources, and dates get cited more than adjectives, and they signal freshness the models favor.
• Build entity authority. Consistent mentions, reviews, and citations across the web teach the model your brand is a credible source to reach for.

That last point is where most teams underinvest. Getting cited by AI is an off-page game as much as an on-page one. In one PR campaign engineered for AI visibility, we chose placements for their odds of being ingested by language models rather than for raw backlinks, and reached a 25% detection rate on LLMs across 565 placements. That is the difference between hoping to get cited and building for it.

If you want the full stack rather than a single tactic, our GEO playbook and our field notes on how to appear in AI Overviews lay out the on-page and authority work side by side.

How to Track AI Traffic Now That SGE Is Gone

You cannot fix what you cannot see. The good news for 2026 is that Google made AI activity more measurable, so you no longer have to guess whether the answer boxes are helping or hurting.

Signal	Where to find it	What it tells you
AI Mode clicks and impressions	Google Search Console totals	Whether AI surfaces send you traffic at all
Citations by prompt	AI visibility trackers	Which questions cite you versus a competitor
Referral sessions	Your analytics platform	Real visits arriving from AI answers
CTR by query type	Search Console, segmented	Where AI answers erode clicks the most

Treat these as a loop rather than a report nobody reads: publish citable answers, watch which prompts quote you, update the pages that get skipped, and double down on the query clusters where you already win. That cycle is the core of ongoing AI search visibility work. If you want to see where AI Overviews and AI Mode already cite you, and where they cite rivals instead, you can book a call and get a free GEO audit back within 24 hours.

Frequently Asked Questions

Are AI Overviews and AI Mode available in every country?

Not everywhere, and not at the same time. AI Overviews launched in the US and have since rolled out to more than 100 countries and many languages, while AI Mode expanded more gradually and reaches fewer markets. Coverage and features vary by region, so check what is live for your target audience rather than assuming parity with the US.

Will AI Overviews replace traditional search results entirely?

No, at least not now. AI Overviews sit above the standard results, and the ten blue links, ads, and other features still appear below them. For commercial, local, and transactional queries especially, Google keeps showing conventional results because people want to compare and choose. The shift is in attention and clicks, not a full removal of organic listings.

Do I need separate content for AI Mode and AI Overviews?

No, one well-built page serves both. Clear answer blocks, real data, and topical depth satisfy the extraction that AI Overviews rely on and the multi-question retrieval behind AI Mode. You are not writing two versions, you are writing one page that answers a full cluster of related questions cleanly enough to be quoted.

Can I stop Google from using my content in AI Overviews?

Partly, but with a heavy cost. The nosnippet and max-snippet directives limit how much of your text Google can display, which can reduce your presence in AI answers, but the same directives also shrink your normal search snippets. Most sites lose more visibility than they protect, so blocking is rarely worth it outside specific legal or paywall cases.

Is SGE the same as Gemini?

No, they are related but distinct. Gemini is Google's underlying family of AI models and its standalone chat app. SGE was the search feature that used those models inside Google Search, and it now lives on as AI Overviews and AI Mode. Think of Gemini as the engine and AI Overviews or AI Mode as the search products built on top of it.