By Mark Buraga, Founder, Growth Engine PH. Last updated 2026-05-31.
Most SEO operators are asking the wrong question about Google I/O 2026’s AI Mode rebuild. They want to know how to “rank” in AI Mode. The real question is whether AI is citing you or mentioning you, and the answer determines what you do this week.
A citation is a source link in AI Mode’s sidebar. A mention is your brand named inside the AI’s synthesized answer body. Citations come from structural retrievability. Mentions come from market consensus. Different mechanisms, different signals, different work. Operators chasing both with the same tactics are wasting half their effort.
This is the operator’s split for the next 90 days. Pick one, structure for it, measure it, then repeat with the other.
Google I/O 2026 rebuilt the front door of the internet, and most SEO dashboards stopped measuring the right thing
Google AI Mode at I/O 2026 replaced the keyword-and-rank search box with a synthesizer that draws from sources outside Google’s top-10 organic results, and the dashboards SEO operators use today do not measure the new visibility surface.
At I/O 2026, Google shipped what it called the biggest upgrade to the search box in over 25 years. Gemini 3.5 Flash became the default model in AI Mode globally. Information Agents started rolling out for AI Pro and Ultra subscribers, operating in the background to surface synthesized answers across the web. Agentic booking expanded to local services. Personal Intelligence in AI Mode reached nearly 200 countries and 98 languages, no subscription required (Google I/O 2026, May 2026).
The mechanical implication matters more than the product list. AI Mode does not show ten blue links. It synthesizes an answer from sources across the web, then shows a sidebar of citations and a body of mentions. The sources it pulls from are not the same set as Google’s top-10 organic results.
The data is now clear on how big that gap is. Only 17% of sources cited in AI Mode rank in the organic top 10 for the same query, down from roughly 76% in mid-2024 (ALM Corp, BrightEdge, and Ahrefs aggregated data, Q1 2026). The page you optimize for organic rank is no longer the page AI Mode is reading. Most operators have not noticed because their dashboards still say everything is fine.
Citations and mentions are two different mechanisms, and operators chasing both with the same tactics are wasting half their effort
A citation is a source link in Google AI Mode’s sidebar, earned by structural retrievability of a single page. A mention is the brand name appearing inside the AI’s synthesized answer body, earned by market consensus across non-website surfaces.
When you ask AI Mode “what is the best home care agency in Maryland,” the response shows two things side by side. The body of the answer might say “Bloomfield Homecare, Visiting Angels, and Care.com all serve the Maryland market.” Those are mentions. To the right, a sidebar lists the URLs the AI drew from. Those are citations.
Citations and mentions are not the same thing, and they are not earned the same way. AI Mode cites sources roughly 76% of the time, but only mentions brand names 38% of the time (cross-platform research, 2026). ChatGPT shows a different pattern, citing 87% of the time but mentioning brands only 21% of the time. Even Google AI Overviews and AI Mode pull from different source pools. Only about 14% of citations overlap between the two surfaces, because they ask different synthesizers to do different jobs.
That means a single page can win a citation in AI Mode without ever being mentioned by name, and a brand can be mentioned by name without ever earning a citation link. They reward opposite work. If you set up your strategy without distinguishing the two, you spend half your budget on tactics that move the wrong meter.
The citation work is structural retrievability, and most operators are still optimizing for the wrong signals
AI Mode’s source sidebar rewards pages with explicit Q-and-A architecture, canonical entity references, clean schema, and dense fact density per paragraph, not page authority or keyword density.
What earns a citation now looks more like what wins a featured snippet than what wins a rank. The page that answers a specific question with a specific answer, in the first paragraph, with the question echoed in the H2 above it, beats the page that buries the answer halfway down a 4,000-word guide. Schema does real work here. FAQPage schema with answer text mirroring the body, Article schema with author and publish date, Person schema with credentials and sameAs links, Organization schema connecting the byline to the publisher. None of this is new SEO, but the priority order changed. Schema went from nice-to-have to load-bearing in 24 months.
The payoff is measurable. Brands cited in AI Overviews earn 35% more organic clicks and 91% more paid clicks than uncited brands appearing only in traditional results, per Seer Interactive’s September 2025 study of 25.1 million organic impressions across 42 organizations. The citation slot is not a vanity badge. It is a click magnet that compounds against rank.
The work, in concrete terms: rewrite your top 10 most-trafficked pages to lead each H2 with a one-sentence quotable answer. Audit your schema and fix the broken or missing types. Add a named author byline with credentials and sameAs links to the named author’s LinkedIn and other public profiles. None of this requires new tooling. It requires sitting with the page and asking: would an AI extract one sentence from this section and quote it cleanly? If yes, the page is a citation candidate. If no, the section is invisible.
The mention work is consensus across non-website surfaces, and that requires a different muscle than on-page SEO
Mentions in AI Mode come from brand names appearing consistently across Reddit, LinkedIn, YouTube, PR coverage, podcast transcripts, and named-expert quotes, surfaces no on-page SEO tactic can influence.
The mention is harder to earn than the citation because the brand has to show up somewhere the AI synthesizes from, and the brand does not control those surfaces. Reddit threads, LinkedIn longform posts, YouTube video transcripts, podcast appearances by the named founder, named-expert quotes in industry articles, press releases picked up by credible publications. The AI weights these signals because they tell it who other people say the brand is, not who the brand says it is.
The data on this is striking. Earned-media coverage distributed across multiple publications increases AI citation rates by up to 325% compared to publishing only on the brand’s own site (cross-publication research, 2026). Review profiles also matter more than most operators realize. Brands with even minimal review profiles, as few as 1 to 13 reviews on a relevant platform, hit a 53.5% AI citation rate (2026 review-platform research).
The work, in concrete terms: identify three platforms where named-author-credible content is feasible for your operation. For a professional services firm, those are typically LinkedIn longform under the founder’s name, YouTube under the firm’s channel, and either a vertical-specific subreddit or a credible industry publication. Set a 90-day calendar that produces two named-author pieces per platform per month. Pitch one podcast per month where the angle would actually land. The work is steady, not heroic. The compounding shows up in months four through six.
Every SEO dashboard built before April 2026 is now incomplete, and that is the measurement architecture problem you need to solve
Traffic-based dashboards (GA4, Search Console, Ahrefs Rank Tracker) do not measure citations or mentions in AI Mode, which is why HubSpot AEO, Profound, Otterly, Athena, and Peec AI all launched in 2025-2026, a measurement gap that no traditional tool fills.
HubSpot disclosed in April 2026 that its own organic traffic dropped 27% year-over-year (HubSpot, AEO product launch announcement, April 14 2026). The company that taught a generation of SEOs how to build the content flywheel watched the flywheel slow down on its own site first. The response is the part worth studying. HubSpot did not write 200 more long-form posts. It did not double the backlink budget. It built a $50-per-month tool to measure something the SEO industry had not given a name to yet: brand visibility in AI answer engines.
The HubSpot speaker on this said the line directly: “your dashboard is just going to be fundamentally lying to you” (HubSpot Marketing Against the Grain, “I Tested Google’s Rebuilt Search: 3 Things to Fix This Week”, May 29 2026). The traffic dashboard kept reporting numbers. The buyer behavior changed underneath the numbers. The gap between what the dashboard said and what was happening in AI Mode grew larger every quarter.
The current tool category is small but real. HubSpot AEO at $50 per month, Profound, Otterly, Athena, Peec AI. Each measures some combination of citations and mentions across AI engines. None of them are required for the first 90 days. Manual capture, 15 prompts per week, 30 minutes of your own time, gets you the baseline you need before the tooling decision matters.
Rand Fishkin says AI search is less than 2% of total traffic, and he is right, but the implied action is wrong
Fishkin’s 2% volume share is accurate, and that does not change the recommendation to learn the citation-vs-mention mechanism now, because mastery compounds independently of share, and operators who wait until 20% will be 18 months behind.
Rand Fishkin made the case in April 2026 that in a zero-click world, traffic is a terrible goal, and that despite the AI search noise, AI usage remains less than 2% of total search traffic (Rand Fishkin, “5 Strategic Features that Predict Survival in the Zero-Click Era”, April 20 2026). The volume number is honest. AI Mode and AI Overviews combined still drive a fraction of the queries Google handles, and operators panicking about a 50% organic CTR drop are reading the headlines, not the data.
The pivot Fishkin’s argument invites, however, is wrong. The recommendation is not to ignore the mechanism until volume hits 20%. The recommendation is to add citation and mention measurement to the stack while organic discipline stays intact. Mechanism mastery does not scale with market share. It scales with reps. Operators learning the citations-vs-mentions distinction at 2% will own the playbook when the surface reaches 20%, because they will have 18 months of data on what works and what does not. Operators waiting will be starting cold.
The investment is asymmetric. Cost to add measurement: 30 minutes per week of manual capture, no tool spend, no team hire. Cost of missing the window: 12 to 18 months of catch-up against competitors who started earlier. The math does not favor waiting.
What this changes in how Growth Engine PH operates, and what it should change in how you do
Growth Engine PH adopted AEO citation and mention audits as a standing weekly deliverable across all retainer tiers effective 2026-05-31, and the operator’s playbook for the next 90 days is to add the same measurement layer before tools are required.
Every Growth Engine PH retainer client, including the trial tier, now receives a weekly AEO citation and mention audit as part of the standard scope. Fifteen prompts per client, locked at onboarding, captured manually, logged in a structured snapshot that feeds the monthly report. The first deliverable in week one is a baseline showing where the client stands today on each of the 15 prompts: zero citations, three competitor mentions, no client mentions. From that baseline, every page-level change and every off-page placement gets attributed against a measurable outcome.
The tier ladder reflects the split. Engine at $999 per month is SEO foundation work plus citation work: page architecture, schema, named-author byline, weekly audit. Engine Pro at $1,499 per month adds active mention-building: LinkedIn longform under the founder’s byline, YouTube content production, podcast pitching, named-expert placement in industry publications. Two different muscles, two different price points, two different timelines to outcome.
For your own operation, the recommendation is the same shape at smaller scale. Pick 15 prompts your buyers would actually type. Capture the AI Mode answer for each, weekly, in a clean browser session. Log citations earned, mentions earned, competitor mentions. Spend 30 minutes a week on this for 90 days. By day 90, you will know which prompts are reachable through citation work and which require mention-building, and the budget conversation in month four becomes specific instead of speculative.
The first audit takes 15 prompts, 30 minutes a week, and zero paid tooling
A useful starting AEO citation and mention audit takes 15 prompts (5 brand-direct, 5 service-intent, 5 competitor-comparison), 30 minutes of manual capture per week, and zero paid tooling.
The 15 prompts split into three buckets of five. Brand-direct prompts test mention coverage on direct brand queries: “is [your firm] legitimate,” “reviews of [your firm],” “who founded [your firm],” “[your firm] pricing,” “[your firm] location.” Service-intent prompts test citation coverage on commercial intent: “best [your service] in [your city],” “[your service] for [your buyer persona],” “how much does [your service] cost,” “what to ask a [your service] firm,” “alternatives to [main competitor] for [your service].” Competitor-comparison prompts test relative mention coverage: “[your firm] vs [main competitor],” “best [service] firms in [region],” “top [vertical] practices,” “who are the leading [service] providers,” “[your firm] alternatives.”
The capture method is mechanical. Open Google AI Mode in a clean browser session, ideally incognito and signed out, with the geo set to your buyer’s primary market. Type the prompt verbatim, exactly as written, without autocomplete edits. Wait for the full answer and sidebar to render. Screenshot the visible area. Count and log the source URLs in the sidebar, flagging any that belong to your domain. Scan the answer body for your brand name, flagging presence and context. Scan for each known competitor’s brand name, flagging presence. Repeat for all 15 prompts. Save the structured result to a folder named with the date, one folder per weekly audit.
The output is two numbers per week per client: citations earned, mentions earned. Plus context: which prompts changed week over week, which competitors are gaining or losing share, which surfaces the AI is drawing citations from. Three months of this produces a dataset no other agency in your competitive set has, because most have not started measuring.
At 90 days, 180 days, and a year, here is what you can expect
At 90 days, operators have measurement they did not have before. At 180 days, the data starts moving decisions about content vs PR investment. At a year, mechanism mastery becomes a moat against late-mover competitors entering the same niche.
The 90-day mark is the measurement baseline. By week 12, you have 12 weekly audits, week-over-week deltas, and a clear picture of which prompts are reachable through page work and which require off-page consensus building. The data is thin but real. The instinct to throw tooling at the problem fades because the manual rhythm is doing the work.
The 180-day mark is where the decisions get sharper. By month six, you have enough trend data to allocate budget between content production and earned media. If the citation count is climbing but mentions are flat, the bottleneck is off-page presence and a PR push or a named-author content sprint is the right move. If mentions are climbing but citations are flat, the bottleneck is page architecture and a schema or quotable-answer rewrite is the right move. The decision stops being a guess and starts being a triage.
The one-year mark is where the moat shows up. The competitor agencies and in-house teams that did not start measuring in 2026 will be one year behind on the learning curve, and the learning curve does not catch up easily because the data takes time to accumulate. HubSpot proved this on itself. The company started building the AEO product in late 2025 because the playbook stopped working on its own site, and by mid-2026 it was selling the measurement layer to the agencies who used to teach others how to win at inbound. A year of measurement mastery is the moat that did not exist before April 2026.
Frequently Asked Questions
What is the difference between an AI Mode citation and an AI Mode mention?
A citation is a source link that appears in Google AI Mode’s sidebar on a given prompt, attributing the answer to a specific URL. A mention is your brand name appearing inside the AI’s synthesized answer body. Citations come from structural retrievability of a single page. Mentions come from market consensus across non-website surfaces like Reddit, LinkedIn, YouTube, PR, and named-expert quotes.
Does ranking in Google organic top-10 still matter for AI Mode citations?
It matters less every quarter. As of Q1 2026, only about 17% of AI Mode citations come from pages also ranking in the organic top 10 for the same query, down from roughly 76% in mid-2024. Optimize for AI Mode citation specifically, not for organic rank as a proxy.
Is AI Overview the same as AI Mode?
No. AI Overview is the AI-generated answer block that appears at the top of classic Google search results. AI Mode is the conversational chat-style interface Google launched at I/O 2026 as the default for many queries. Only about 14% of citations overlap between the two surfaces, because they pull from different source pools.
How do I measure mentions if no tool does it natively?
Two paths. Manual: pick 15 prompts you would want your brand to surface in, type them weekly into AI Mode incognito, log whether your brand name appears in the answer body. Tooled: HubSpot AEO at $50 per month, plus Profound, Otterly, Peec AI, and Athena all track brand mentions in AI Mode with varying depth and price. The manual path is sufficient for a first 90 days.
If AI search is only 2% of traffic, why invest now?
Volume share is not a useful proxy for when to start learning a mechanism. Operators who learn the citation-vs-mention split at 2% will own the playbook when AI search reaches 20%. Mechanism mastery compounds with reps, not with market share. The investment is low (30 minutes per week of manual measurement) and the upside is asymmetric.
What pages on my site are most likely to be cited by AI Mode?
Pages with explicit Q-and-A architecture, dense fact density per paragraph, clean schema (FAQPage, Article, Person, Organization), canonical entity references, and a named author byline. Long-form posts with these structural signals outperform short landing pages even when the landing pages have higher conversion rates.
Where do mentions come from?
Reddit threads, LinkedIn longform posts, YouTube video transcripts, podcast appearances, press coverage on credible publications, named-expert quotes in industry articles, and review profiles on category-specific platforms. Earned-media coverage distributed across multiple publications increases AI citation rates by up to 325% compared to publishing only on your own site.
Should I hire a PR firm now to chase mentions?
Not yet. Spend the first 90 days measuring which prompts your brand is missing on, then decide whether the gap is best closed by content (citations) or earned media (mentions). The cheapest first move is to seed your own LinkedIn and YouTube presence with named-author content under the founder’s byline. A PR firm engagement is a Month 6+ decision, not a Day 1 decision.
Stop asking how to rank in AI Mode. Start asking what you want, citation, mention, or both, and structure the next 90 days around the answer.
Sources: Google blog, Search at I/O 2026. HubSpot, AEO product launch. HubSpot Marketing Against the Grain video. Seer Interactive citation click study, September 2025. ALM Corp, AI Overview Citations Top-10 Drop. Ahrefs, 38% of AI Overview Citations Pull From The Top 10. Search Engine Land, AI Overviews drive CTR drop. PPC.land, HubSpot AEO tool launch. Rand Fishkin, “5 Strategic Features that Predict Survival in the Zero-Click Era,” SparkToro, April 20 2026.