Key takeaways
- Gemini 3.5 Flash launched at Google I/O 2026 on 19 May and is the default model powering Google Search AI Mode and the Gemini app, available to all users rather than only paid subscribers.
- Google reports the model runs about four times faster than comparable frontier models, often at under half the cost, and beats Gemini 3.1 Pro on coding and agentic benchmarks.
- AI Mode reached 75 million daily active users, making the AI answer a primary destination, not a side feature.
- Ahrefs found AI Mode and AI Overviews cite the same URLs only 13.7% of the time, so agencies must optimise for each surface separately.
- Only 38% of pages cited in AI Overviews now rank in Google’s top 10, down from 76% in mid-2025, breaking the old link between ranking and AI visibility.
- Generative Engine Optimisation (GEO) becomes a paid service line: structured content, clear entities, brand mentions across the web, and freshness now drive citations.
- Agencies that sell answer-first content, citation tracking, and entity clarity gain new revenue; those selling rank reports alone face shrinking value.
SEO, internet data analytics – artistic impression. Image credit: Growtika via Unsplash, free license
Gemini 3.5 Flash will push SEO agencies away from chasing blue-link rankings and toward earning citations inside Google’s AI answers. Google now runs this model as the default engine behind AI Mode and the Gemini app for billions of people worldwide, so the box that summarizes a query before any link appears is the surface clients most want to win. Agencies that treat that answer box as the new front page keep their clients visible; those that keep optimizing only for position one lose ground.
The second change is structural. Research from Ahrefs shows that Google’s AI Mode and AI Overviews pull from almost entirely different source lists, overlapping on cited URLs just 13.7% of the time. A page can dominate one surface and never appear in the other, which means agencies now manage two parallel visibility channels instead of one ranking report. The work shifts from tracking ten positions to making content easy for a fast model to read, trust, quote, and attribute.
What is Gemini 3.5 Flash?
Gemini 3.5 art. Image credit: Google
Gemini 3.5 Flash is Google DeepMind’s fastest frontier-class model, released at Google I/O 2026 as the first member of the Gemini 3.5 family. Google described the family as “combining frontier intelligence with action,” pointing to a model built to plan and complete multi-step tasks rather than answer a single prompt. Flash carries a one-million-token context window, accepts text, images, audio, video, and PDFs, and produces up to roughly 65,000 output tokens per response.
The headline trait is speed paired with capability. Google says Flash outperforms its previous flagship, Gemini 3.1 Pro, on coding and agentic benchmarks while running about four times faster than rival frontier models. The model works with Google Antigravity, the company’s agent-first development platform, to run several subagents at once, and it powers Gemini Spark, a personal agent that acts on a user’s behalf. Gemini 3.5 Pro sits in internal testing and is expected next month.
| Specification | Gemini 3.5 Flash |
|---|---|
| Release date | 19 May 2026 (Google I/O 2026) |
| Context window | 1,000,000 tokens |
| Max output | ~65,000 tokens |
| Inputs | Text, image, audio, video, PDF |
| API input price | $1.50 per million tokens |
| API output price | $9.00 per million tokens |
| Cached input | $0.15 per million tokens |
| Where it runs | Gemini app, AI Mode in Search, Gemini API, Antigravity, Android Studio, Gemini Enterprise |
For agencies, the pricing matters as much as the speed. At $1.50 input and $9.00 output per million tokens, Flash sits roughly 25% below Gemini 3.1 Pro while beating it on the tasks that touch search. Cheaper, faster inference lets Google serve AI answers to everyone, which is why the AI box now greets nearly every searcher instead of a tested minority.
Why Gemini 3.5 Flash matters for search
When a faster, sharper model becomes the default behind Google Search, the answer arrives before the links do, and it arrives more often. AI Mode already counts 75 million daily active users, and Google shipped Flash straight into search products on launch day rather than holding it back for months. That deployment pace tells agencies something practical: model upgrades now reach clients’ search results almost immediately, so visibility can move week to week, not quarter to quarter.
Flash also reads pages more thoroughly. It reasons across multiple sources, weighs which ones agree, and assembles a single response with citations attached. A page no longer competes only for a rank; it competes to be one of the sources the model trusts enough to quote. That standard rewards clarity, structure, and corroboration far more than keyword density.
How Gemini 3.5 Flash changes the daily work of SEO agencies
From rankings to citations
The clearest change is the target. An agency used to report client positions for a keyword set; now it must report whether the client appears inside the AI answer and which sources Google chose. Ahrefs found that only 38% of pages cited in AI Overviews still rank in Google’s top 10, down sharply from 76% in mid-2025. Ranking first no longer guarantees a citation, and a page on the third results page can still surface in the answer. Agencies that keep selling rank dashboards alone will struggle to explain flat traffic to clients whose answers are being written above the links.
Two AI surfaces, two source lists
AI Mode and AI Overviews look similar but behave like separate engines. Ahrefs studied 540,000 query pairs and found the two cite the same URLs only 13.7% of the time, rising to just 16.3% across the top three citations. Despina Gavoyannis, Senior SEO Specialist at Ahrefs, noted that nine times out of ten the two systems “agreed on what to say” while pointing readers to different pages. For an agency, that splits one optimisation job into two: a brand visible in AI Overviews appears in AI Mode only about 61% of the time, so each channel needs its own audit and its own content plan.
| Signal (Ahrefs, Sept 2025 US data) | AI Mode | AI Overviews |
|---|---|---|
| Same URLs cited for a query | 13.7% overlap | |
| Semantic similarity of answers | 86% | |
| Average domains cited per query | 9.2 | 7.7 |
| Entities mentioned per answer | 3.3 | 1.3 |
| Wikipedia citation rate | 28.9% | 18.1% |
| Answers with no citation | 3% | 11% |
The table points to a tactical read: AI Mode quotes more sources, names more brands, and almost always cites someone, which makes it the more reachable surface for clients who publish authoritative, well-structured pages.
Faster answers, fewer clicks
Speed has a downstream cost for traffic. When Flash returns a complete answer in a fraction of a second, more searchers read it and stop, and click-through to the underlying pages keeps sliding. Agencies feel this as a widening gap between impressions and visits. The honest conversation with clients moves from “how much traffic did we win” to “how much of the answer did we shape,” because a cited brand earns trust and consideration even when the click never lands.
Content built for extraction
Flash favours pages it can lift cleanly. That rewards a direct answer near the top of the page, plain headings that match real questions, defined entities, and facts a model can verify against other sources. Long-form depth still helps, but only when the page also serves a quotable, self-contained passage. Agencies that rewrite client libraries around extractable answers give the model fewer reasons to skip them.
New services SEO agencies can sell
The shift opens revenue rather than closing it. Generative Engine Optimisation, often shortened to GEO, sits next to traditional SEO as a paid line covering structured data, entity-clear architecture, and third-party citation building. Agencies can package AI-answer visibility audits, brand-mention monitoring across the web, schema and freshness programmes, and separate playbooks for AI Mode versus AI Overviews. Because YouTube and other formats now appear in citations that never rank as web pages, multi-format authority work, including video and structured Q&A, becomes a sellable deliverable too.
Flash also lowers the cost of the agency’s own production. The model writes valid JSON-LD schema, drafts dozens of meta descriptions, maps internal links, and outlines briefs around real search intent. Teams that fold Flash into their workflow ship more client work per hour, which protects margins as reporting grows more complex.
What SEO agencies should do next
The practical path starts with measurement. Agencies should track citations inside both AI Mode and AI Overviews, treat them as separate scorecards, and tie client value to share of answer rather than rank alone. Next comes content surgery: place a clear, quotable answer high on every priority page, sharpen entity definitions, and keep facts current so the model reads the page as fresh and reliable. Building brand mentions across trusted third-party sites raises the odds that Flash sees the client as a source worth quoting, since corroboration across pages is what the model weighs.
None of this retires classic SEO. Crawlability, site speed, intent-matched content, and authority still decide whether a page qualifies to be read at all. Gemini 3.5 Flash raises the bar on top of those basics: it asks agencies to make every client page easy to read, easy to trust, and easy to cite, across two AI surfaces that now decide what most searchers see first.
If you are interested in this topic, we suggest you check our articles:
- Beyond Bard: The Power of Google Gemini
- Google Traffic: AI vs Human Content – Navigating Expectations, Reality, and Future Prospects
- Gemini AI Chrome Integration: Enhanced Web Browsing Features
- The Future of Autonomous AI
- Writesonic: AI Writer for Standout Content Creation
Sources: Google Blog, Google AI for Developers, Ahrefs, Search Engine Journal, Android Authority, eWeek, DesignRush
Written by Alius Noreika
