Creators Now Need to Persuade Algorithms, Not Just People

For a decade, creator marketing ran on two numbers: reach and engagement. A new report from Digiday’s Future of Marketing Briefing suggests a third input has quietly muscled its way into the brief — whether a creator’s content gets cited by AI models when someone asks a chatbot what to buy.

What the agencies are seeing

According to Digiday, agencies across the market are fielding the same question from clients. At Reign Maker, AI visibility has been the first or second topic raised with every client for six months straight. Ogilvy says a handful of its more advanced clients already have explicit KPIs around discoverability, credibility and culture.

Tinuiti offers the sharpest measure of the shift. Six to eight months ago, AI visibility was roughly a tenth of the factors it weighed when building creator strategy; now it’s closer to a quarter, and the topic comes up with around 60% of its client roster. “It’s front and center right now,” Crystal Duncan, evp of brand engagement at Tinuiti, told Digiday.

Jellyfish reports that more than 90% of its marketers now rank AI’s impact on discovery and sales in their top three concerns. Its “Share of Model” tool runs large batches of prompts against major LLMs and tracks which sources get cited across tens of thousands of results, then feeds that back into creator briefs.

The finding that breaks the old playbook

Here’s the part that should make every influencer marketer sit up. Go Fish Digital’s Lauren Lyster told Digiday that citations aren’t tracking with the metrics brands optimise for: “It’s not necessarily built on engagement rates or follower counts.” Long-form content is outperforming Shorts in citations, she said — short-form rarely gets picked up at all.

The Now Agency’s benchmark is domain depth held consistently over years. Co-founder Gabe Feldman told Digiday the agency wouldn’t back a creator who spent five years on beauty and skincare and is now talking about coding. Under those rules, a 100,000-follower specialist can outweigh a five-million-follower generalist.

Why this is a psychology story, not just an SEO one

Language models are, in effect, a machine approximation of a very human heuristic: authority bias. People trust the person who has demonstrably done the thing, repeatedly, in one lane. Models are trained on that same signal — consistency, specificity, verifiable claims — because that’s what human-written text treats as credible.

Which means the machines aren’t asking for something alien. They’re asking for the thing consumers already rewarded, just measurable at last. Notably, Digiday cites data showing only 23% of U.S. shoppers rank AI assistants as trusted shopping sources — behind family and friends, professionals, retailer reviews and influencers. The chatbot is a router to trust, not a replacement for it.

What to actually change in your briefs

Dept’s approach, as reported, is instructive. It checks whether a creator already appears in AI answers for the category, then layers that on top of the usual filters — category authority, credible content, audience trust. The change shows up in the brief itself:

• Hand creators a reference list of exact product names and verifiable claims to script from
• Structure chapter markers around the real questions people ask, not plain timestamps
• Supply a clean transcript rather than letting models parse error-filled auto-captions
• Weight the creator mix toward voices already cited in your category
• Treat long-form as the citation asset; short-form as the attention asset

Dept’s vp strategy Angela Seits stressed the content should still be natural and in the creator’s own voice — the structure behind it is what changes.

The caveat worth holding onto

Rate cards haven’t moved. Creators are folding citations into media kits as a differentiator rather than a price increase, and Tinuiti execs expect that to shift only in another six to 12 months, once brands understand what a citation is worth.

There’s also a credible over-engineering risk here. Ogilvy’s Charlie Coney noted the opacity of it all — LLMs take time to train and the mechanics aren’t public, so there’s guesswork involved. Content that is helpful, entertaining and ownable was good creator work long before anyone measured it against a model. The machines just made the case for depth harder to ignore.

Source: Digiday