A marketing director opens ChatGPT, types the exact prompt a buyer would use to find a company like hers, and watches the same five logos surface again. Her brand is not one of them. She asks Gemini. Different list, mostly the same shape at the top.
She asks Perplexity. The lead names shuffle but the pattern holds. Somewhere in the shortlist sits the competitor she outsells on G2, outranks on Google, and outspends on paid. The model hasn’t noticed.
This isn’t a bug in one tool. It’s the shape of generative answers, and it’s now reshaping which companies get considered at all. The decisions below are the ones every marketing leader outside that short list has to make before next quarter’s pipeline review.
Decide Whether the Concentration Is Real Enough to Act On
The first decision is whether to treat brand concentration in AI answers as a measurable fact or a vibe. The research says fact. A recent academic analysis of generative search found what the authors call “Big Brand Bias”: major brands account for a disproportionate share of mentions, with a small handful of names dominating the head of the distribution in category after category. Soft drinks, airlines, insurance, software — the shape repeats.
One nuance worth absorbing: the leaders aren’t identical across engines. ChatGPT, Gemini, Perplexity, and Google’s AI Mode each build their shortlists from different source mixes, so the top five in one tool isn’t the top five in another. Treat concentration as directional, not deterministic. It tells you the ceiling is low and the competition for a slot is fierce.
What it doesn’t tell you is that the slot is unwinnable.
If you’ve been waiting for the numbers to stabilize before building a plan, stop waiting. The pattern is stable enough. What’s unstable is which engine your buyer uses next week, and that argues for working on the inputs all of them share — your brand’s own site, third-party mentions, and the structured references models pull from — rather than optimizing for one chatbot’s quirks.
For a deeper look at the mechanics, Digital.Marketing’s write-up on the AI citation concentration problem covers how links, mentions, and web longevity feed the shortlist.
Decide What a Lost Mention Actually Costs You
Pipeline math is where this stops being an SEO conversation and starts being a CFO conversation. The honest read is that AI referral traffic is still small relative to classic search — under 2% of B2B referral traffic in one analysis of 25 million sessions — but it’s growing fast, and the visitors who do click tend to arrive further along in their decision. A smaller stream of warmer leads is not nothing.
So the decision is how to price the miss. Two ways to frame it:
- Direct referral loss. Count the sessions and demo requests you aren’t getting because the model named someone else. Small today for most B2B categories. Larger for consumer and local.
- Consideration loss. Count the deals where you never made the initial list because the buyer’s first research pass happened inside a chatbot. Harder to measure, almost always bigger.
Budget only against the first number and you’ll underfund the work. Talk only about the second and finance will tune you out. Bring both.
Decide Where to Spend: Owned Site, Earned Mentions, or Structured References
The shortlist draws from three inputs, and you have to decide which one to push on first. Picking all three at once is how programs stall.
- Your own site. It carries more weight than people assume — a brand’s own domain can account for a majority of the citations an engine uses when answering a branded query. Fix the pages a model would quote: comparisons, specs, pricing logic, clear category language.
- Third-party mentions. Reviews, roundups, trade coverage, and the long tail of industry sites that get crawled into the training and retrieval mix. This is where challengers close the distance on incumbents, because incumbents stop earning fresh mentions once they’re famous.
- Structured references. The pages a model trusts as neutral ground. Independent databases, standards bodies, well-maintained category explainers. Being named inside one of these is worth more than ten brand-authored blog posts.
For most challengers, the second lever moves fastest. The first takes a quarter to show. The third takes a year, but it compounds.
Decide Who Owns This Before Q1 Planning Closes
The reason most companies are losing ground here is organizational, not technical. SEO sees it as a backlog item. PR treats it as a byproduct of coverage.
Product thinks it’s a schema problem. Nobody runs the number at the monthly review.
Pick one owner. Give them a budget line, a short list of prompts to monitor, and a quarterly target for share-of-answer in the categories that matter. The owner doesn’t have to be a new hire — the head of SEO, the head of comms, or the demand gen lead can hold it. What it cannot be is “all of the above,” which is the arrangement that got you here.
The five brands at the top of the answer didn’t win a secret auction. They won by being referenced, over and over, across the open web long before the models were trained. Catching up is slower than paid search and faster than brand building. It starts with deciding it’s someone’s job.
