The GEO market is booming, usage is flat: the gap nobody looks at

by | Aug 26, 2026 | GEO

The market is on fire. Fifty-five percent of marketing leaders now have a dedicated budget line. Agencies are hiring specialists. Tracking tool vendors are raising money. One agency recommends a pilot budget worth one and a half to two times the existing search budget.

Everyone infers from this that usage is following.

It is not following. At least not in the large Western markets, and not where it can be measured cleanly.

The problem is not that generative AI is a flash in the pan. The problem is that budgets are being sized against an assumed adoption curve, while the available data describes a plateau. This article puts the two curves side by side.

What the Reuters Institute measures

The Digital News Report is published every year by the Reuters Institute for the Study of Journalism, attached to the University of Oxford. The 2026 edition, out on 16 June, covers forty-eight markets and is the fifteenth run of the survey.

Provenance marking: an annual academic survey, large sample, published methodology, no commercial interest in the result. That is the highest level of evidence available on self-reported usage.

The global figure first. Ten percent of respondents use AI chatbots to get news, against seven percent the previous year. Among the under-thirty-fives, that rate rises to sixteen percent.

Growth, then. Three points in a year.

And here is the line nobody quotes. That growth does not exist everywhere. The United States, the United Kingdom, France and Germany record no increase over the year.

Four of the largest Western markets. Zero progress. The global growth is coming from somewhere else.

The honest caveat, which limits this argument

The limit has to be stated immediately, otherwise this article commits the fault it is denouncing.

The Digital News Report measures chatbot use for getting news. It does not measure general ChatGPT usage, which is growing strongly. Nor does it measure whether your customers are asking an AI which vendor to choose, which software to compare, or which plumber to call.

So a reader can legitimately object that news is a special case, and that commercial research may well follow another curve. That is a valid objection, and no public data of this quality settles it for any of these four markets.

What this survey establishes is narrower, and sufficient: on the only usage measured finely and longitudinally, with academic methodology, there is no growth in 2026. Any projection assuming exponential adoption in your market has to rest on something other than data.

The other curve: the traffic actually received

The second anchor point is the traffic generative platforms actually send to websites.

An analysis published on 6 July 2026 covers 13,770 domains, including 1,215 client domains for the traffic component, representing 3.3 billion sessions. Provenance: a marketing tool vendor, methodology described, United States scope, data from May to September 2025, ten sectors.

Traffic referred by AI platforms accounts for 1.08 percent of visits. It is growing at roughly one percent per month.

One percent. Not thirty-five, not fifteen. One.

That figure circulates in distorted form more or less everywhere, often inflated by a factor of thirty through confusion with market share between chatbots, a metric whose denominator has nothing to do with this one. If someone tells you AI represents a third of your potential traffic, ask for the denominator.

The two curves, side by side

Indicator	Value	Trend	Source
Marketers with a GEO budget line	55 percent	Rising sharply	Survey, 205 executives, March 2026
Pilot budget recommended by agencies	1.5 to 2x the search budget	Rising	Same survey
Chatbot use for news, worldwide	10 percent	+3 points in a year	Reuters Institute, 48 markets
Chatbot use for news, United States, United Kingdom, France, Germany	Flat	No increase	Reuters Institute, 2026
Share of traffic referred by AI platforms	1.08 percent	+1 percent per month	1,215 domains, United States
Click on a link displayed in a generated summary	~1 percent	Flat	Pew Research, ~900 users

The budget follows one curve. Usage follows another. The gap between them is not an opportunity, it is a premium paid on an anticipation.

What argues the other way

Three objections deserve to be put, because they are serious.

General usage is growing, even if informational usage has plateaued. Chatbots are gaining users. Nothing says the plateau observed on news will repeat itself on commercial usage, which does not have the same trust dynamics.

One percent growing at one percent a month is not nothing. At that rate, the order of magnitude changes within a few years. Sizing to zero would be as wrong as sizing to thirty-five.

And the main effect may be invisible. A June 2026 paper, resting on a panel joining real browsing and AI conversations from the same users, measures that a brand recommendation lifts Google searches for that brand by 4.3 points. That effect never travels through an AI referrer: your analytics will attribute it to branded organic. Referral traffic therefore understates the real impact, structurally.

Those three objections do not restore the adoption curve you are being sold. They indicate that the right number is neither one nor thirty-five, and that it is specific to your market.

How to measure at your own company, instead of believing an average

The only serious way to size this budget is to measure your own exposure. Four indicators are enough, and three are free.

Isolate in your analytics the sessions whose referrer belongs to a generative platform. You will get your own equivalent of the 1.08 percent, and it will probably differ: technical sectors and software sit well above the average, local commerce well below.

Track branded search in Search Console, on a rolling twelve months. That is where the indirect effect shows up, with a lag.

Query the engines yourself on your ten category prompts, several times, varying the wording. Count the share of answers where you are named, and the share where a competitor is named in your place. A single measurement is worth nothing: generative visibility is a distribution, not a point.

And compare the cost of the proposed engagement to the revenue genuinely attributable to that exposure. If nobody knows how to run that calculation, it is not an investment, it is a bet.

What you do tomorrow morning

Ask your vendor, or yourself, which adoption data the proposed budget rests on. Not an intuition, not a line from a conference stage: a dated source, with a sample size.

If the answer is a global projection, remember that the United States, the United Kingdom, France and Germany are the four markets where the reference survey measures no progress this year.

Size against what you measure, not against what the market anticipates. The market has been anticipating for eighteen months, and it has been billing the whole time.

Sources

<strong>LaFactory</strong> measures AI visibility with a published protocol: repeated measurements, paraphrases, control group. No guaranteed placement, ever. Contact us to scope an audit.