A 50,000-keyword study shows Google AI Mode now runs ads on a meaningful share of commercial searches, and paying for that slot does almost nothing for whether your brand gets cited. This is the full picture: what's actually happening in AI search right now, and the concrete steps that earn a citation instead of buying a placement.
Google AI Mode is now a real ad surface, not an experiment, and paying for it buys you a slot, not a citation. A study of 50,032 commercial keywords, all specifically chosen because they're eligible to trigger an ad, found a text ad appeared in 29.45% of responses, rising to 53.56% for keywords with a cost-per-click of $10 or more. In the same dataset, advertisers were cited as a source in only 12% of the keywords they were paying on, and at the exact-URL level that overlap collapsed to under 2%. Roughly 85% of those same advertisers didn't even appear in organic results for the keywords they were bidding on. Three separate systems, paid placement, organic ranking, and AI citation, running in parallel, and doing well in one tells you almost nothing about the other two.
Worth being precise about what that ad number actually measures, since it gets flattened in most coverage. It's the ad rate among keywords selected specifically for their ad eligibility, not a claim about how often ads show up across every kind of search. It also swings hard by category, from over 70% of ad-eligible keywords in some niches down to under 3% in others, heaviest in lead-generation categories, rarest in informational and YMYL topics. Your actual exposure to this depends far more on which vertical you're in than on the headline figure.
So what's actually happening in AI search right now, beyond the ad question. Google AI Mode passed a billion monthly users in 2026, and both AI Mode and the AI Overviews that show up inside regular search results now run on the same underlying Gemini model, Google moved to Gemini 3 as the default for AI Overviews starting this January. AI Overviews themselves now appear on a large and growing share of Google searches, and the effect on regular organic clicks is significant, industry tracking suggests a meaningful drop in click-through on queries where an AI Overview appears, with the click volume that remains concentrating hard onto whichever handful of sources actually got cited. Being cited isn't a nice-to-have anymore. For a growing share of queries, it's the only visibility that survives.
It also isn't one system to optimize for. ChatGPT pulls its web results from Bing's index. Claude pulls from Brave's. Google's AI Overviews and AI Mode run entirely on Google's own index. The same question, asked in three different places, can return three completely different sets of cited sources, because each engine is drawing from a different pool underneath. A GEO strategy built around only one of these will look successful in the tool you're checking and be invisible everywhere else your customers are actually asking.
Here's what actually earns a citation, and it's remarkably consistent across engines rather than being a secret per platform. Answer the question directly and completely in the first sentence or two of a section, before any windup or context-setting, that's the passage most retrieval systems pull first. Phrase your subheadings and section leads as the actual question someone would type or ask an AI, not a clever label, because that's what the retrieval layer is matching against. Structure matters more than length, comparison tables, numbered steps, and clean short paragraphs get pulled over long unstructured prose even when the prose says the same thing. And back specific claims with a real number, a named source, or a first-hand detail, generic assertions get skipped over for the version of the same point that's more checkable.
Freshness is a real, separate signal, not a courtesy. Google's own guidance points to updated-date signals mattering specifically for anything time-sensitive, pricing, statistics, trends, features. A page that's visibly current tends to get preferred over an identical page with an old date attached, especially in categories that move fast. That has a direct implication for content like this piece: an AI citation earned today on a fast-moving topic isn't permanent. Industry tracking on commercial queries puts citation churn somewhere in the range of 40 to 60 percent, meaning a meaningful share of cited sources get swapped out over time as competing pages update, rank shifts, or an engine simply rotates its sources. Treat citation as something you check and refresh on a schedule, not something you earn once and leave alone.
None of this requires special markup that doesn't exist elsewhere. Google has been direct on this point, there's no dedicated schema exclusively for AI Overviews, the lever is the same one that's always mattered, genuinely useful, well-structured, technically accessible content, with the extraction-friendly structure above layered on top of solid fundamentals, not replacing them.
A few of the exact questions worth having direct answers to, since these are close to how people are actually asking: does buying a Google AI Mode ad help you get cited? No, per this study, paying for the placement and being cited are close to statistically unrelated. Do you need to rank number one to get cited? Not necessarily, being cited is now more about being the clearest, most extractable, most trustworthy source on the specific question than about holding the top organic position. Can you track whether you're actually appearing in AI Mode? Partially, Google Search Console has counted AI Mode appearances in its Performance report since mid-2025, and manual spot-checking in a private browser window against your top commercial keywords remains the most direct way to see exactly what's being cited right now, since AI citation data isn't yet exposed as a clean report the way rankings are.
The practical version of all of this, this week: pull up your top five commercial keywords in AI Mode yourself. Note which ones show an ad. Separately, note who's actually cited. If those two lists barely overlap, and per this data they usually will, that's the signal to stop budgeting for AI visibility and paid placement as if they're the same line item, and to start treating citation-earning structure as its own, ongoing piece of work.
If your paid spend and your actual AI visibility might be telling two different stories, that's worth finding out before the next budget cycle.
Book a Growth Audit →© 2026 The Chill Pill