Three stories today cover agency AI briefs, B2B websites that have to do more than rank, and The RealReal's run-in with Meta's shopping agent.
Today's stories:
1. Agency AI runs on briefs, and most briefs are too thin
2. Winning AI search is step one, and your website is step two
3. The RealReal tightens bot defenses after Meta's Muse agent hoards carts
Agency AI Runs on Briefs, and Most Briefs Are Too Thin
Kelly Hayes, head of agency growth at Adswerve, argues in Demand Gen Report that most agencies already have AI tools and haven't fixed what goes into them. She cites Deloitte's State of AI in the Enterprise report, where 37% of organizations report using AI only at a surface level, with little or no change to their underlying business processes. Yes, this is a story about briefs in a series called The Brief, and no, I didn't plan that.
A brief used to be a few lines written for a person, like "millennial homeowners, premium positioning, avoid discount language," and a media buyer filled in the rest with judgment. An AI system reads those same lines literally and fills every gap on its own, which often produces generic or off-brand work. Hayes points to Google's AI Brief, which lets advertisers steer AI Max campaigns in plain language with guidelines for messaging, search matching, and audiences.
Her advice is to tell the AI what the brand won't say, describe audiences by behavior and intent, define success in business terms like customer lifetime value, and keep revising the brief as results come in. The data has to be ready first, since behavioral audiences need person-level identity resolution across devices and conversions need a value finance recognizes. Hayes also wants a senior brief owner who can push back on clients, and she suggests starting with one account where a thin brief is costly.
Winning AI Search Is Step One, and Your Website Is Step Two
Scott Gillum, founder and CEO of Carbon Design, writes in MarTech that showing up in Google, ChatGPT, or Perplexity doesn't guarantee a sale. The website still has to help buyers understand the offer, trust it, and act on it. That job starts early, since 6sense's 2024 Buyer Experience Report found buyers finish nearly 70% of the buying process before they talk to a seller, and 81% already have a preferred vendor at first contact.
Visits are short, with Databox putting the median average session duration for B2B companies at 78 seconds. Digital Commerce 360 reported that only 35.7% of B2B buyers rate their experience on seller websites as somewhat or very positive. Gillum says vague claims like "industry-leading" and "AI-powered" make buyers doubt you, while research-heavy buyers want specifics on methodology, security, and implementation.
Gillum's fix is cognitive-load testing across four areas, which are comprehension, orientation, credibility, and action. In plain terms, can a visitor tell what you sell, can they find their way around, can they trace your claims to real customers and sources, and are there next steps beyond "Contact sales" on every page? Vague language and scattered facts can also lead AI systems to summarize a site inaccurately, so question-based headings, consistent product names, and claims placed next to their evidence help on that front too.
Read more: https://martech.org/why-winning-ai-search-isnt-enough-to-win-b2b-buyers/
The RealReal Tightens Bot Defenses After Meta's Muse Agent Hoards Carts
The RealReal says it is updating its bot defenses to make it harder for AI agents to shop on its site. The change follows a Verge report that Meta's Muse agent could exploit the retailer's item reserve system. Items sit on hold for 20 minutes once they're in a cart, and The Verge got Muse to re-add listings as soon as each hold expired, which blocked other shoppers from buying them for hours or longer.
After the story ran, chief technology officer Luke Friang said in an email that The RealReal had addressed the Muse use case. Its bot defense configuration now makes it much more difficult for all agents to log in, add items to cart, and check out, though agents can still surface listings and direct Muse users to product pages. Friang added that the company will keep strengthening protections as AI-powered shopping evolves.
The Verge points out that a person will only re-add designer clothing to a cart so many times, and an agent doesn't care. Meta said on its Q3 earnings call that shopping is the most common use case for Muse, with an average purchase size over $200, and Amazon has blocked agents from its site. The RealReal is building its own, with Ask TRR launching to all 45 million members after a pilot. More on this in my posts on agentic AI in ecommerce and agentic AI in retail.
Read more: https://www.theverge.com/news/1018713/meta-muse-agentic-shopping-ai-the-realreal
That's it for today.
Follow me on LinkedIn and BlueSky, and X
FAQs
What is Google's AI Brief?
AI Brief is a Gemini-powered feature in Google's AI Max for Search campaigns. Advertisers write plain-language guidelines for messaging, search matching, and audiences, and they can preview sample assets and searches before a campaign goes live.
How do you write a good brief for AI ad tools?
By spelling out what the brand won't say, describing audiences by behavior and intent, and defining success in business terms like customer lifetime value.
Is winning AI search enough to win B2B buyers?
Most often, no. 6sense found buyers finish nearly 70% of the buying process before they talk to a seller, so the website often gets judged before sales is involved.
What happened with Meta's Muse and The RealReal?
The Verge got Muse to keep re-adding RealReal listings to a cart whenever the 20-minute hold expired, which blocked other shoppers from buying them for hours or longer. The RealReal's chief technology officer says the company has since made it much more difficult for all agents to log in, add items to cart, and check out.