10,000 prompts. Five engines. Fifteen categories. Here's what the flywheel actually looks like.
In the categories my firm tracked this year, a small number of brands captured the majority of citations across ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews. Not a plurality. A majority. The pattern is more concentrated than any share-of-voice or organic-search leaderboard I've seen in twenty-three years of running 5W.
That's the headline of The State of AI Search 2026 — the annual we just published at 5wpr.com. Sixty-two pages. Five engines. Fifteen-plus consumer and B2B categories. Ten thousand buyer-intent prompts. A year of citation tracking, scored against a fixed rubric.
The report has ten findings. Read all of them. But the argument compresses to three sentences.
One. The market is concentrating faster than the leaders realize.
Category after category, the top-cited brand doesn't just lead — it compounds. The brand cited first inside AI answers captures far more buyer-research mindshare than its raw citation rate predicts. The engines default to the entity they know best. The buyer takes the first answer. The first answer feeds the next citation. The flywheel spins.
This is not share-of-voice. Share-of-voice was a percentage of mentions. Citation share is a percentage of answers. The two are not close. When the answer is one paragraph, being in that paragraph is the only market share that matters.
Two. Citation decay is real, measurable, and faster than the CFO expects.
Brands that paused earned media and structured-content investment lost measurable citation share inside months. Often before any traditional metric moved. Site traffic still looked fine. Brand tracker still looked fine. The citation curve had already broken.
AI visibility is not a launch campaign. It's an operating cost. The brands winning citation share are the brands publishing, being covered, being referenced, and being cited on a weekly cadence — quarter over quarter, year over year. The moment they stop, the engines re-weight toward whoever kept going.
Three. Local is the largest under-cited category in the economy.
HVAC. Plumbing. Dental. Funeral services. Car wash. Home services. Across every local-services category we tested, most operators were not cited at all when buyers asked the engines for a recommendation. Not ranked low. Not present.
The category worth trillions of dollars annually is running on Google Maps and word of mouth. Ask ChatGPT for a plumber. Watch the answer. The businesses being recommended are not the ones with the largest revenue in the market. They're the ones with the strongest third-party citation footprint — Reddit threads, local review sites, community forums, editorial mentions. Nobody's optimizing for that yet. The first movers in local AI visibility will own their metros for a decade.
The pattern under the findings.
Each engine cites differently. ChatGPT leans on structured reference. Claude leans on editorial press. Perplexity leans on community and primary sources. Gemini leans on the Google graph. AI Overviews lean on schema density. Single-engine optimization leaves citation share on the table — because the buyer is not on one engine.
Reddit is a top-tier citation input across every category with active community discussion. Not a distribution channel. A citation input. Brands that treat Reddit as spam surface lose citation share to brands that treat it as an owned engagement channel. This is uncomfortable for legacy PR. It's also settled by the data.
Schema markup is table stakes. The absence of it is a critical gap. The presence of it is not a differentiator. The next-frontier signal is structured citation density across third-party sources — how often, how consistently, and with what surrounding entity context your brand appears in the corpus the engines actually retrieve from.
Earned media remains the strongest single GEO input we measured. Not because journalists are gatekeepers. Because independent third-party validation is what the engines have been trained to trust.
Why we published this as an annual.
The AI Visibility Index Series covers categories one at a time. Beauty. Legal Tech. Defense & Aerospace. Consumer Electronics. Luxury. Creators. The Index Series answers who's winning in this sector. The annual answers what's the pattern across every sector at once.
The methodology mirrors the Index Series and extends it to a full annual sample. Every prompt, every scoring rubric, every confidence interval is disclosed in the report. Any communications team can replicate it. Any competitor can challenge it. That's the point.
The 2026 GEO playbook inside the report is prescriptive. Ninety-day, six-month, twelve-month frameworks — segmented by company size, category, and starting position. Not theory. Operating instructions.
The one line I want the industry to hold.
The buyer is now asking a machine. The machine is picking a small set of brands to name. The set is concentrating, not expanding. The brands not in it are not competing — they're absent. Absence is the new losing.
AI Communications is a mix of journalism, psychology, and engineering — and the audience is now the machine.
Read the report. Download it at 5wpr.com/research/state-of-ai-search-2026. It's free. It's sixty-two pages. It's the reference now.