Digiday published a by-the-numbers roundup on 7 August 2026, written by Kimeko McCoy, collecting survey data on how marketers are funding visibility inside AI search. The lead figure comes from Fractl: marketers now route roughly 24% of their search and content budgets to AI visibility work. Four other vendors' surveys in the same roundup describe where that money sits and how little of it is being measured.
One caveat before the numbers. This is five separate studies collected by one outlet, not a single piece of research. Fractl, Similarweb, 10Fold, Scribewise and Semrush each measured different things, so every figure below is attributed to the vendor that produced it. The source for all of them is Digiday.
Fractl: about a quarter of search and content budget now goes to AI visibility
Fractl's data, reported by Digiday on 7 August 2026, puts the average allocation at roughly 24% of search and content budgets. Adoption is wider than that average suggests. Fractl found that 82% of marketers have allocated at least some budget to AI visibility work, while 18% allocate nothing at all. At the committed end, 43% of marketers spend more than 20% of that budget on it.
The spread matters more than the mean. A market where almost one in five spends zero and more than two in five spend over a fifth has not settled on a standard yet.
Fractl: performance and SEO teams are carrying the spend
Fractl also split the allocation by department, and the split is uneven. Performance marketing commits 32% of its budget and SEO commits 31%, the two largest shares. Brand commits 21% and content commits 20%, the two smallest. The ordering suggests AI visibility is being paid for as an acquisition line rather than a publishing line. The table below gives the split as Fractl reported it.
| Department | Share of budget committed to AI visibility (Fractl) |
|---|---|
| Performance marketing | 32% |
| SEO | 31% |
| Brand | 21% |
| Content | 20% |
Similarweb: AI search leads discovery, then loses its edge at purchase
Similarweb data in the same Digiday roundup found that AI search accounts for 35% of initial product discovery, against 13.6% for traditional search. The gap narrows further down the funnel: at the purchase stage the split is 24.3% for AI search against 22.1% for traditional search. Discovery is where the difference is large. By the time someone is ready to buy, the two channels are close to level.
10Fold: B2B marketers rate answer engines above organic search
10Fold's survey, cited by Digiday on 7 August 2026, found that 52% of B2B marketers rate AI answer engines their most effective distribution channel, against 29% for organic search. The same survey found that 41% have only 25% to 49% of their content optimised for AI discovery. The channel is being rated first and supplied last.
Scribewise: most brands cannot tell whether any of it is working
Scribewise data in the roundup found that 67% of marketers report their brand appears in AI answers less often than they want. Alongside that, 71% do not track competitive share of voice in AI answers, and 70% do not monitor brand sentiment there. Roughly seven in ten are therefore spending against a target they have no reading on, which makes the 67% a feeling about visibility reported mostly by people not measuring it.
Semrush: how many sources an AI answer actually pulls
Semrush figures in the same Digiday piece give the average number of sources cited per answer as 15.4 for ChatGPT, 11.4 for Google AI Mode and 3.3 for Gemini. A Gemini answer citing three sources is a far narrower opening than a ChatGPT answer citing fifteen, so anyone planning generative engine optimisation work is aiming at three targets of very different sizes.
What this means for Thai marketers
None of the five surveys reports Thai data, and the Digiday roundup contains no Thailand-specific figures. What follows is reasoning from the benchmark, not something the source states.
The benchmark itself is the usable part. If peers elsewhere are moving something like a quarter of search and content budget into AI visibility while seven in ten track neither competitive share of voice nor brand sentiment, then matching the spend is the expensive way to copy them and measuring is the cheap way to get ahead of them. A Thai brand can establish where it currently appears in AI answers for its own category terms, in Thai and in English, for a fraction of what a year of unmeasured work costs. That baseline also answers whether the spend is warranted at all, which the Fractl numbers cannot.
The department split reads locally too. If performance and SEO fund this globally, a Thai organisation that parks the work with content alone is funding it from the smallest budget in the building. The mechanics of AI SEO overlap heavily with the technical and content groundwork already behind SEO in Thailand, so in most cases this is a reallocation question rather than a new discipline.
Frequently asked questions
Did one study find that marketers spend 24% of budget on AI visibility?
No. The 24% figure belongs to Fractl alone. Digiday collected it in a roundup alongside separate surveys from Similarweb, 10Fold, Scribewise and Semrush, each measuring something different. Treating the set as one study would misstate all five.
Is there any Thai data in these numbers?
No. None of the five surveys reported by Digiday breaks out Thailand, and the roundup makes no Thailand-specific claim. Any read on the Thai market from these figures is inference, and should be labelled as such.
Which teams are actually funding AI visibility work?
Performance marketing and SEO, according to Fractl, which put their commitments at 32% and 31% of budget. Brand at 21% and content at 20% commit the least. Fractl did not report why the split falls that way.
Do I need to move budget right now?
The source does not say, and it makes no recommendation. The numbers support a narrower point: 71% of marketers do not track competitive share of voice in AI answers, so a measurement baseline is a first step that costs less than a budget shift.
Why does the number of sources per answer matter?
Because it sets how many brands can be cited in a single answer. Semrush put the average at 15.4 sources for ChatGPT, 11.4 for Google AI Mode and 3.3 for Gemini, so the same content faces very different competition depending on the engine answering the question.
Benchmarks are most useful once you know where you stand against them. If you want a reading on how often your brand appears in AI answers for your own category terms, Relevant Audience can help you set that baseline.
