Marketing has always had one audience: people. This year it quietly gained a second one. Braze’s 2026 Customer Engagement Review puts consumer use of AI agents for brand interactions at 19% today — and projects 46% by the end of the year. Agents that research, compare, shortlist and increasingly transact on a person’s behalf.
The projection may land lower. It doesn’t matter much. Even the conservative version means a meaningful share of your next hundred prospects will first meet you as a machine-readable summary, assembled by software that has never seen your brand film.
What the numbers actually say
- 19% → 46% within the year. That’s the projected jump in consumers using agents for brand interactions — research and comparison first, purchases increasingly included.
- 27% refuse to share any data with an agent. Trust, not capability, is the constraint. The brands that win agent traffic will be the ones that are verifiable, not the ones that shout loudest.
- The perception gap is real. 93% of marketing leaders believe AI helps them understand customers accurately; 53% of consumers agree. The gap between those numbers is where credibility goes to die.
What an agent can’t be sold
An agent doesn’t feel urgency banners. It isn’t charmed by hero videos. It reads structure: what you offer, for whom, at what terms, with what proof. It cross-checks your claims against reviews, directories and whatever else it can reach. Ambiguity, which human visitors forgive, reads to an agent as absence.
The second audience doesn’t watch your brand film. It reads your pricing page.
What to fix first
- Say what you do in extractable form. Every service needs a page that answers who, what, how and on-what-terms in plain text — not spread across a deck you email on request.
- Mark it up. Organization, Service and FAQ schema turn your claims into data an agent can lift without guessing.
- Make your facts agree with each other. Name, offering, locations, claims — identical across site, profiles and directories. Agents treat contradictions as risk.
- Invest in verifiable proof. Reviews and third-party mentions are what an agent uses to rank you against the shortlist. That footprint compounds slowly — start now.
We wrote about the local-search version of this shift in local visibility in the AI era — the same logic now applies to every considered purchase, including choosing a marketing partner.
Sources
Questions we get on this
We’re B2B — is agentic commerce really our problem?
Yes, with a different verb. Consumer agents buy; business agents shortlist. When a buyer asks an assistant for “embedded marketing support in Switzerland” or “clinic marketing specialists”, the agent assembles a comparison from whatever it can read and verify. If your service pages are vague about what you do, for whom, and on what terms, you’re not in the comparison.
What’s the single first fix?
One honest, plainly written page per service: what it is, who it’s for, how the engagement works, and how pricing is structured. Add structured data (Organization, Service, FAQ schema) so the same facts are machine-readable. Agents reward extractable clarity — the same thing human buyers reward, with less patience.
Do we need to buy an AEO or GEO tool for this?
Mostly no. The work is editorial and structural, not tooling: clear pages, consistent facts across your site and profiles, real reviews, schema markup. A tool can monitor whether assistants cite you; it can’t create the substance they’d cite. Start with the substance.
