Only 42% of advertisers can see their creator agency fees, and that number explains the week

Four things landed within 48 hours this week: a Digiday briefing on creator agency fees nobody can audit, a publisher coalition releasing a technical standard for counting what AI systems do with content, publishers turning AI visibility into a paid product, and MIT Sloan arguing that service providers survive in-house AI only by absorbing work clients can't check themselves. The common mechanism is disclosure of units. Each development forces an intermediary to say what it counted, who counted it, and how much of the money passed through to whoever did the work.

The thesis for CMOs: visibility, attributionattributionA framework for assigning credit to the touchpoints that contributed to a conversion, so you can measure which channels and interactions actually drive results.View full definition → and intermediary trust are all being rebuilt on the same primitive, which is event-level evidence of what a middleman actually did. The teams that write that requirement into contracts this quarter will set the price of creator, media and AI visibility work for the next three years. The teams that wait will buy bundled lines with a margin baked in and no way to compute it.

Only 42% of ISBA members can see their creator agency fees

Digiday's Future of Marketing Briefing, published 2 October 2026 by Seb Joseph and Krystal Scanlon, put a number on something marketers have been muttering about for two years. William Bradley, a senior director at Responsible Marketing Advisory, polled members of ISBA's media leaders group and creator forum after the trade body asked his consultancy to build a framework for managing creator work: 42% said their creator agency fees were fully transparent, 21% said partially, and 37% said they were unsure. Bradley's summary of the findings was that the practice has "outgrown governance."

The pricing evidence is sharper than the poll. Marketing management consultant TrinityP3 argued that a bundled creator deal can leave as little as 35% of the budget for talent, against as much as 60% when the same work is billed line by line. ANA figures cited in the same briefing put the average agency take at about 30% of influencer spend, with only 39% of agreements showing transparent terms. Digiday also notes the obvious caveat, which is that the consultants raising the alarm would like to be paid to fix it. The pattern is familiar: programmatic had its reckoning a decade ago, principal media faces the same accusations now, and agencies appear to be rolling talent fees and their own margin into a single creator line clients can't pick apart.

Hypothetical, arithmetic only, using TrinityP3's published range. Take a £2m annual creator budget. At the bundled end, 35% reaching talent is £700,000. At the unbundled end, 60% is £1.2m. The £500,000 difference is not a saving you negotiate, it is a disclosure you either have or don't. Nothing in that calculation requires a new agency; it requires an invoice with a pass-through column.

What to do: reopen your creator statements of work before the 2027 planning cycle closes and demand talent fee, usage rights, production, platform fees and agency margin as separate lines. This is the samediscipline you already apply to media rebates and principal buying, applied to a channel that grew too fast for your procurement team to keep up.

What is SPUR's content telemetry standard, and does it bind AI labs?

It binds nobody. It is a voluntary open specification, and as Digiday reported this week, no AI company has accepted the invitation to help govern it. SPUR, the Standards for Publisher Usage Rights initiative, released its content telemetry standard and invited OpenAI, Anthropic, Google, Meta and Microsoft to shape implementation through a new invitation-only AI Licensing Advisory Board, according to technical lead Alex Springer. Springer expects the board to reachreachThe number of unique people exposed to your message in a given period. Unlike impressions, reach counts each person once, no matter how often they see it.View full definition → about 20 members with a first meeting planned for October; no AI company had accepted at publication, a Google spokesperson said only that the company engages with SPUR and other associations on various topics, and the others did not respond.

The substance is the schemaschemaA schema is the formal blueprint that defines how data is structured, named, typed, and related within a database, file, or message.View full definition →. The framework breaks AI content use into five measurable events, content retrieved, grounded, cited, displayed and engaged, and creates a common format for reporting those interactions back to publishers. SPUR was founded in March 2026 by the BBC, the Financial Times, The Guardian, Sky, The Times of London and MediaHaus, with the AP joining as first US founding member alongside 30 publisher members and six affiliates.

Why a CMO should care about a publisher licensing fight: those five verbs are the first credible vocabulary for AI attribution that was not invented by a tool vendor trying to sell you a dashboard. "Cited" and "displayed" and "engaged" are different events with different commercial value, and almost every AI visibility product sold today collapses them into one score. Nothing to implement. Watch two things: whether any frontier lab joins that advisory board before the end of the year, and whether your own AI visibility vendor can report against those five event types. If it can't, you are buying an opinion.

Future's GEOGEOThe practice of making your brand and content visible and citable inside AI-generated answers from tools like ChatGPT, Gemini and Perplexity.View full definition → product has more than 30 clients and renewals

Digiday's Media Briefing of 1 October 2026 reported that generative engine optimisation has moved from publisher experiment to rate card. A year ago GEO barely registered on publishers' rate cards; now Future's GEO product has more than 30 clients and secured renewals, Ziff Davis says it has multiple GEO clients, and one publisher exec said its offering is already making money, though the figures don't yet add up to a clear picture of market size. Publishers haven't been deterred by AI companies pushing back: Perplexity told Digiday last month it would block its agents from scraping Time's site after the publisher began serving ads to its agents in markdown files, yet Perplexity's agents are still scraping Time and recognising those markdown ads, according to a person familiar with the analytics.

Read that against the creator fee story and the structure repeats. Your media partner now sells the visibility it also measures, with no third party counting the events. That is the same conflict as an agency that books the creator, sets the fee and reports the result.

Separately, the vendor layer around this is loud. Semrush, which sells SEOSEOSearch Engine Optimization: the practice of improving your pages' natural (unpaid) rankings in search engine results pages to attract more organic traffic.View full definition → and AI visibility software, published a run of pieces this week on answer engine optimisation, llms.txt, AI visibility audits and agentic SEO workflows. Treat those as product marketing, not market evidence. On llms.txt specifically the independent picture is unflattering: an Ahrefs analysis of 137,210 domains found 97% of llms.txt files received zero requests in May 2026 (Ahrefs also sells SEO tools), an SE Ranking study of 300,000 domains put adoption at roughly 10.13%, and Google's Gary Illyes has confirmed Google doesn't support the file and isn't planning to, while John Mueller compared it to the discredited keywords meta tag.

What to do: before signing a publisher GEO deal, require the unit of measurement, the query set, the refresh cadence and a holdout of brand terms you don't optimise, so you can tell lift from drift. The discipline is the one you already use formulti-touch models that can't be audited end to end.

SharkNinja says creators make 70% of its content

At Shoptalk Fall in Nashville, three retailers put operating numbers on creator programmes. SharkNinja's chief brand and experience officer Michelle Crossan-Matos called creators the company's "magic sauce" and said 70% of the content on SharkNinja is created by creators. Crossan-Matos said content about the Ninja Creami generated 1 billion impressionsimpressionsThe total number of times an ad or piece of content is displayed, regardless of clicks. Each display counts as one impression, even to the same person.View full definition → last year. SharkNinja makes 25 new products a year and builds a creator plan for each. Bob's Discount Furniture showed a graphic at Shoptalk indicating the four couples featured in season two of its programme have a combined reach of some 12 million people, with organic reachorganic reachThe number of people reached without paid promotion, through content shared naturally via feeds, search, and word of mouth.View full definition → and engagement running 50% higher than season one at the same point last year. The same Digiday piece lists this year's expansion: Aerie launched a creator programme for its most avid fans, Gap Inc. started an employee creator programme, Ugg worked with art students on creative and product pages, Target launched Club Target and Target Ambassadors, and Kohl's began paying creators commission on every sale.

Now set that against the supply side. Digiday reported on 1 October that the creator economy loudly denounced generative AI content earlier this year after the Sora 2 backlash, with YouTube removing AI slop accounts and Instagram's Adam Mosseri publishing a memo on ranking for originality, but AI has since crept into nearly every corner of the creator economy, and the harder question is how much of the creative process creators will hand over. CapCut carries a suite of AI editing tools, Adobe Firefly is billed as an all-in-one creator studio, and YouTube is adding agentic tools including video A/B testingA/B testingA/B testing is a controlled experiment that compares two versions of something (A and B) by splitting traffic randomly to learn which performs better on a chosen metric.View full definition → and conversational editing. Lucy Kruszynski, global head of brand at influencer agency Buttermilk, told Digiday the industry is in a different place morally on AI and that it's harder to be binary about it now.

If 70% of your content comes from creators and an unknown share of that is now AI-assisted, your disclosure obligations, your brand safety review and your talent fee logic all rest on something you don't measure. What to do: add an AI use declaration to creator contracts covering ideation, editing, voice and likeness, and decide which of those you pay full rate for.

The underrated signal this week is MIT Sloan on in-house AI

MIT Sloan Management Review published "How to Outcompete Your Client's AI" on 29 September 2026, opening on real estate investment firm Alturas Capital Partners, whose in-house lawyers moved lease work away from outside counsel using generative AI, saving hundreds of thousands of dollars and compressing work that once stalled for weeks. Drawing on WPP, Moody's, A&O Shearman and Thomson Reuters, the authors set out three moves providers can make to stay the preferred choice: improve unit economics, become easier to buy from, and absorb the quality-assurance burden clients would otherwise carry. Thomson Reuters is their example of the third move, with CoCounsel Legal built on Westlaw and its decades of curated case law and 35 million legal classifications.

Everyone read that as a consulting story. It is an agency pricing story. The creator fee argument and the AI capability argument converge on the same point: once a client can produce a passable version of the deliverable in-house, a bundled fee with an undisclosed margin stops being defensible, and the only margin that survives is attached to work the client cannot verify or maintain alone. The MIT Sloan authors make the counterpoint that bringing work in-house looks cheap until someone has to check every output, fix mistakes, rewrite prompts as models change and clear compliance, costs that accumulate until they rival the original fee. That is the honest case for keeping an agency, and it is a case your agency should be making in units, not adjectives.

The decision rule, short enough for a whiteboard: for every intermediary line item, name theunit, the counter and the pass-through. What is being counted (a cited event, a creator post, an impressionimpressionThe total number of times an ad or piece of content is displayed, regardless of clicks. Each display counts as one impression, even to the same person.View full definition →), who counts it and can you reproduce the count without them, and what percentage of that line leaves your building and reaches the person or platform doing the work. Any line that fails all three is margin, not cost, and should be priced as risk.

What to do on monday

Pull your three largest creator invoices from the past year and try to fill the unit, counter and pass-through columns yourself. Whatever you can't fill becomes the agenda for the next partner review.

Send your creator agencies a one-page fee disclosure request modelled on the ISBA framework work, asking for talent fee, usage, production, platform fee and margin as separate lines, with a response deadline inside two weeks.

Ask your AI visibility vendor or publisher GEO partner to mapmapUsing software to automate repetitive marketing tasks and campaigns, enabling personalisation at scale across channels like email, web, and social.View full definition → its reporting to SPUR's five events (retrieved, grounded, cited, displayed, engaged) and tell you which ones it cannot observe.

Add an AI use declaration clause to the creator contract template and brief legal on where likeness and voice sit.

Put a standing item on your agency QBR: which deliverables could we now produce in-house at acceptable quality, and what is the partner absorbing that we would otherwise have to check ourselves.

The 42% figure is not a scandal number, it is a procurement gap. Close it on creator spend this quarter, because the same question is about to arrive on every AI visibility invoice you sign, and the vocabulary for answering it was published this week by a group of publishers rather than by anyone selling you software.