Google AI Mode and the May 2026 search update changed how B2B buyers research software before contacting sales or booking demos.
Published May 20, 2026
TL;DRGoogle's May 2026 I/O update made Gemini 3.5 Flash the default model for AI Mode globally, accelerating the gap between traditional rankings and AI citations.Top-10 rankers accounted for 76% of AI Overview citations in mid-2025 but only roughly 38% by early 2026, meaning strong rankings no longer predict AI visibility.Winning in AI search requires three simultaneous games: ranking, citation, and agent-source-trust. Most B2B SaaS teams optimize for one and ignore the other two.Passage retrieval and information consistency across independent sources drive AI citations more than backlink count.Add self-reported attribution fields and HubSpot segments today to track AI-referred leads as a distinct pipeline source.
This guide covers what Google announced at I/O 2026, how the Gemini 3.5 Flash update changes retrieval, and a three-bet quarterly roadmap to build content that gets cited. For a broader treatment of where SEO and Answer Engine Optimization (AEO) overlap and diverge, our piece on whether AEO is different from SEO is a good starting point.
Google's May 2026 I/O announcements confirmed that AI Mode is no longer experimental. Google made Gemini 3.5 Flash the default model globally for the Gemini app and AI Mode in search, added multimodal capabilities, integrated autonomous agent workflows, and placed ads directly into AI Overview responses. For B2B SaaS, this means your primary search surface now synthesizes answers rather than returning a ranked list, which changes how buyers find and evaluate software vendors before they ever book a demo.
Google built Gemini 3.5 Flash for speed without sacrificing quality. As the announcement stated, "you no longer have to trade quality for latency." For content strategy, this matters because faster retrieval means content structured for clean passage extraction competes more effectively. AEO research on 2 million citations and 10,000 pages suggests that structural clarity is a significant factor you can control.
Dense retrieval systems outperformed keyword-matching by 9 to 19 points on top-20 passage retrieval (Karpukhin et al.). If your pages aren't structured as independent answer blocks, the model moves past them. Multimodal capabilities add another dimension: buyers can now upload a software architecture diagram and ask which tools integrate with that stack, or run voice queries about pricing mid-commute. That means evaluation decisions can happen without your text-only pages entering the picture at all. Add integration-specific content, architecture compatibility pages, and structured comparison tables that a model can extract for these query types.
Autonomous AI agents now research software vendors on behalf of buyers. An agent tasked with "find the top three incident management tools for a 200-person SaaS company" queries multiple sources, synthesizes a recommendation, and delivers it without the buyer viewing a search results page. My full guide to winning AI search for B2B SaaS covers how to map these agent-driven research journeys and what content structure each stage requires.
Structure your content to answer the follow-up questions agents generate as part of query fan-out, not just the top-level query. Pricing, integration depth, migration complexity, and support response time all belong inside your content, because agents pull from whatever source answers those sub-questions most cleanly.
Google now integrates ads directly into AI Overview responses. Paid placements appear alongside organic citations, which makes organic citation the higher-trust signal. Buyers who see a brand cited organically inside an AI-generated answer treat it differently from a paid placement in the same response: the citation reads as editorial validation. For your quarterly plan, prioritize organic citation investment over AI Overview ad spend until you have baseline visibility data showing where paid will actually move pipeline.
Winning in AI search requires mastering three simultaneous games: ranking, citation, and agent-source-trust. B2B SaaS companies that plateau in AI visibility are usually winning one game and ignoring the other two. My video on why SEO is not AEO breaks down where the retrieval systems diverge and why that gap matters for tactical priorities.
Traditional web search still matters. Buyers still click through to vendor websites, and strong rankings generate traffic that converts. The error is treating this as the only game. Your domain authority and indexing history tell Google how to rank your pages in a results list. They don't tell Gemini which passage to extract and synthesize into an AI answer. Those are different scoring systems, and they require different optimization inputs. Technical SEO foundations, including clean site architecture, fast load times, proper indexability, and internal linking, remain the prerequisite before any citation-layer work has a chance to compound.
Game two is passage retrieval. This is where the CITABLE framework operates. CITABLE has seven components designed specifically for LLM extraction:
Before incident.io worked with us, their CMO described the gap clearly:
"Before Discovered Labs, we were using homegrown LLM prompts, without a clear strategy for what to optimize for or exactly how best to structure content." - Tom Wentworth, CMO at incident.io
After applying the CITABLE framework, incident.io's AI visibility moved from 38% to 64%, closing the competitive gap with PagerDuty.
Game three is information consistency across the open web. Google's AGREE research confirms that LLMs weight claims appearing consistently across independent sources more heavily when grounding answers. That shifts off-page strategy from "acquire do-follow links" to "keep the same accurate claim about your product live across Reddit, industry publications, comparison content, and your own site."
Our analysis of 144,000 ChatGPT citations from 2025 found Reddit appeared in only 0.35% of visible ChatGPT citations but occupied roughly 27% of ChatGPT's internal search slots during query processing. That gap between what's visible and what's influencing answers is exactly why a links-only view of off-page misses most of the retrieval surface. Our Reddit marketing service builds this information consistency as part of the broader off-page motion, not as a standalone tactic.
AI Mode changes where pipeline originates not just where impressions land. Buyers who form vendor preferences inside an AI-generated answer often skip the organic click entirely, which means ranking data and session counts no longer tell the full pipeline story.
Stable Google rankings can coexist with a declining AI citation rate. When that happens, traffic holds steady because ranked pages still get clicks from buyers who scroll past AI Mode, but MQL volume drops because the AI-informed buyer segment never reaches your site. Sova Assessment, a B2B SaaS client, is the clearest example from our own client work: after shifting from purely ranking-focused investment to AI search optimization, organic search became Sova's number one pipeline channel, contributing more than 50% of pipeline. Rankings and citation rate are measuring different things, and reporting only one of them creates a blind spot that shows up as a conversion problem.
Add these four metrics to your monthly executive report alongside traditional SEO KPIs:
Traditional metric | AI-era equivalent | What it measures |
|---|---|---|
Keyword rank | Citation rate | % of monitored buyer queries where your brand appears in AI answers |
Organic sessions | AI-referred sessions | Sessions with an AI platform referrer or attributed source |
Impressions | Share of voice | Your citations vs. competitors across a defined query set |
Our AEO performance metrics guide covers how to pull these numbers into a monthly report your CEO reads as a pipeline story, not a dashboard dump.
B2B buyers use AI assistants across multiple research sessions before contacting sales. Each session is a citation opportunity, and most of these sessions don't leave a standard referral cookie in GA4. Set up a "how did you hear about us?" field on your demo request form with AI assistant options listed explicitly, including ChatGPT, Perplexity, Google AI Mode, and Claude. Train your BDRs to ask the same question on every discovery call and log the answer as a HubSpot contact property. That self-reported data surfaces AI referral that analytics tools miss entirely. Our AI visibility tracker covers the citation-side measurement, while the attribution steps below handle the CRM side.
These three bets form your quarterly roadmap. They sequence from highest immediate leverage to longer-build compounding work, so you see signal within two weeks while the structural investment accumulates in the background.
Your pricing, product feature, and comparison pages are the highest-leverage starting point because they already have domain authority and indexing history. Add a 2 to 3 sentence BLUF at the top of each section, break content into 200 to 400 word extractable blocks with clear heading questions, and add FAQs to every money page. Internal linking between product pages and supporting content helps AI systems map your entity relationships explicitly.
Run each money page through our free AEO content evaluator to score it against the CITABLE framework. Prioritize pages that rank on page one but aren't being cited: Google trusts the page but LLMs aren't selecting passages from it. That gap is your fastest win because the domain authority is already there. Our guide to AI SEO tools covers how to run this audit across a full content library.
Map the priority buyer queries your ICP actually asks an AI assistant. If your brand appears in a fraction of them, identify the gaps and prioritize by pipeline value, not search volume. Ship one direct-answer article per gap per week and measure citation rate after 60 days. Our research on what drives AI citations, drawn from 2 million citations and 10,000 pages, gives you the evidence base for which content attributes move citation rate most.
One well-structured piece per week on the right queries typically outperforms sporadic publishing on high-volume keywords by a wide margin.
"We're working with Discovered Labs and have achieved excellent results. Highly recommend the team's delivery — they consistently put out a higher volume of content than the norm." - Verified user review of Discovered Labs
AI systems surface inaccurate claims about pricing, features, and integrations. You'll find it harder to correct incorrect information in AI answers than to fix a bad G2 review because there's no moderation flow you can flag. The fix is information consistency: publish the accurate claim on your site, corroborate it on Reddit, and confirm it in comparison content and review summaries. My Reddit strategy video covers how to build this motion systematically without violating platform rules.
Attribution is where most teams fall short. Without it, you can't justify the investment or tell a credible pipeline story to the CEO.
Use our AI visibility tracker to monitor your citation rate and share of voice across ChatGPT, Claude, Perplexity, Gemini, and Google AI Mode. We documented the measurement issues most tracking platforms have so you know what to verify before trusting any platform's numbers in an executive report. Run the tracker against the same priority buyer queries you mapped in bet two to keep measurement anchored to pipeline-relevant terms.
For HubSpot attribution, set up this four-step workflow:
AI-referred session segment. Track the MQL-to-opportunity conversion rate for this segment separately to establish whether AI-referred leads deliver a lower CAC than other organic sources.
Share of voice is your citation count divided by total citations across a monitored query set, compared to your top competitors. If your brand appears in 12 answers across 50 monitored queries, competitor A in 23, and competitor B in 7, your share of voice is 12 out of 42. That's 28.6%. Run this monthly: a rising share of voice is the leading indicator of AI-referred pipeline growth.
Combine your AI-referred HubSpot segment with your standard pipeline attribution model. If AI-referred contacts convert to opportunity at a materially higher rate than other organic contacts, as our anonymized 90-day case study shows is achievable, you can justify the investment as a lower-CAC channel. In your monthly CEO report, frame it as "AI-referred contacts convert to opportunity at X rate, delivering an estimated $Y in pipeline this quarter." That shifts the conversation from "is AEO working" to "how fast should we scale this channel."
The question to ask any agency is concrete: can they show you a methodology for measuring citation rate, do they have in-house AI/ML engineers building retrieval tooling, and can they run a live visibility audit on your domain right now? Many agencies that added AEO to their service list in 2025 rebadge existing content and link-building work without changing the underlying retrieval mechanics. The mechanism, not the label, is what determines whether citation rate moves.
Tom Wentworth at incident.io described the gap between having a clear retrieval strategy and improvising one:
"I have recommended you to multiple peer CMOs. There are large organizations like Hubspot and Ramp who have dedicated teams to work on large projects like AEO. For everyone else (except my competitors) there's Discovered Labs!" - Tom Wentworth, CMO at incident.io
We start every engagement with an AI visibility audit using our proprietary AI visibility software. If you want a structured proof window before committing to a retainer, the AEO Sprint at $7,995 (approximately €6,995) delivers 10 CITABLE-optimized articles, a full visibility audit, schema implementation, and a 30-day action plan. The Starter retainer at $7,995 per month (month-to-month, no annual lock-in) covers up to 20 articles per month using the CITABLE framework, off-page consistency work, structured data, visibility tracking, and a dedicated team of four. Book a call and we'll tell you honestly whether we're a fit.
Google AI Mode is a dedicated AI-powered search interface that uses Gemini to synthesize multi-step answers rather than returning a ranked list of links. AI Overviews appear inline within standard search results, while AI Mode is a separate experience. As of May 2026, Gemini 3.5 Flash powers both globally.
Structure your content using the CITABLE framework: 2 to 3 sentence answer-first openings, 200 to 400 word extractable sections, FAQPage and Article schema, and verifiable claims with cited sources. Build the same accurate product claims consistently across Reddit, review platforms, and independent publications to establish information consistency across the sources LLMs use for grounding.
Initial citations appear within 1 to 2 weeks of publishing CITABLE-structured content. Meaningful share of voice improvement across priority buyer queries takes approximately three months of consistent publishing. Full optimization across all three surfaces (web search, citations, training data) takes four to six months.
Combine three signals: a self-reported "how did you first hear about us?" field on your demo form with AI assistant options listed, BDR discovery call notes logged as a HubSpot contact property, and a HubSpot segment built on referral sources from AI platforms (perplexity.ai, chatgpt.com, etc.). Three data sources triangulate to a reliable estimate that's defensible to finance and to your CEO.
No. Top-10 rankers accounted for 76% of AI Overview citations in mid-2025 but only roughly 38% by early 2026. Ranking well supports indexability but doesn't determine which passages AI systems extract and synthesize. Passage structure and information consistency are the additional inputs that drive citation selection.
The percentage of monitored buyer queries where your brand appears in an AI-generated answer. Calculated as (your citations / total queries monitored) x 100.
Your citation count compared to competitors across a defined query set. If you appear in 12 answers and competitors combine for 30, your share of voice is 28.6%.
The process by which LLMs extract specific text blocks from indexed content to synthesize answers. Optimized by structuring content into 200 to 400 word independent sections with clear answer-first openings.
The degree to which the same factual claim about your product appears across independent sources: your site, Reddit, reviews, and industry publications. LLMs weight consistent claims more heavily when grounding answers.
Discovered Labs' seven-component content structure designed for LLM passage retrieval: Clear entity and structure, Intent architecture, Third-party validation, Answer grounding, Block-structured for RAG, Latest and consistent, Entity graph and schema.
Sessions where the originating referral source is an AI platform (perplexity.ai, chatgpt.com, etc.) or where a contact self-reports an AI assistant as their first discovery channel.
Discover more insights on AI search optimization
Most AEO dashboards report rate moves without uncertainty bounds. Here's the math and the prompt-set, variance, and trend tests every measurement should pass.
Read articleIs AEO or GEO different to SEO? This article covers how the difference in technologies impact the tactics and priorities.
Read articleGoogle AI Overviews does not use top-ranking organic results. Our analysis reveals a completely separate retrieval system that extracts individual passages, scores them for relevance & decides whether to cite them.
Read articleOur team analyzed network traffic from Google AI Mode in January 2026. The capture included 547 Google flows and over 1,300 total requests during AI Mode sessions. The findings paint a clear picture of how Google is preparing to monetize AI-generated search results.
Read article