You can now run performance-based ads on ChatGPT

Ten weeks in, OpenAI ads have gone from velvet rope to open bar.

Are you ready to order: OpenAI has started offering cost-per-click (CPC) bidding to select advertisers within its ad pilot, Digiday reports. And the pricing? $3 to $5 per click.

If this seems accessible compared to the initial release… that’s because it is.

Since February 2026, OpenAI dropped minimum spend from $250,000 to $50,000 and CPMs have cooled from $60 to around $25.

All this signals ChatGPT’s quick shift to self-serve. Brand advertisers are happy paying per impression. Performance marketers aren’t. They want to pay for clicks, and now they can.

Of course, this brings up the platform comparison questions to the fray:

• Does a ChatGPT click carry Google Search-level intent?
• How does it stack up against Meta’s high-volume, lower-intent traffic?
• Do your current measurement tools even give you a clean read?

On that last point: Measurement is still shaky, so any early results could still be directional.

If you’re one of the pilot advertisers, low minimums and thin competition may be worth a small test before the crowd shows up.

Your audience is already using ChatGPT to discover and evaluate products, barriers to entry are low, and the competition is thin.

In some ways, it does feel like a happy hour deal.

Pinterest proves its ROI, Google eavesdrops smarter, and Microsoft makes LinkedIn targeting more inclusive

For years, Pinterest was the mood board.

Now it’s the cash register too: A report across 17 matched market tests found that 82% of Pinterest campaigns saw positive incremental returns, with 76% hitting 1.5x ROI or higher.

Pinterest catches shoppers before they’ve made up their minds, which is what separates awareness spend that actually converts from awareness spend that just looks good.

If those numbers hold consistently, Pinterest earns a real line in your media mix.

Google’s listening, just like your FBI agent: Smart Bidding’s conversion signal has switched from call duration to AI-qualified calls, and recording is now on by default for eligible accounts.

Call Details reports also get AI-generated summaries and intent hashtags like #HighIntent or #ConsultationScheduled, so you can see exactly what’s driving the machine.

If your conversion setup hasn’t changed but your performance data suddenly looks different, this is probably why. Check your settings now.

Microsoft Advertising isn’t the little sibling your mum makes you play with anymore: LinkedIn profile targeting now spans 10 job seniority levels and custom company lists.

And a separate audience builder lets you describe your ideal customer in plain English and generates the full segment for you.

Together, they bring LinkedIn-level targeting precision to a platform that typically runs at lower CPCs.

Sounds like someone’s life might get just a little easier…

What biases do AI agents use to choose your product?

Your next customer might not have a heartbeat.

Research out of Columbia and Yale, covered by the Science Says team, maps the early buying biases of today’s AI agents.

The buyer is no longer always human: Tools like ChatGPT’s Agent Mode, Google’s Buy for Me, and Amazon’s Rufus can now search, evaluate, and purchase on a user’s behalf.

We’ve spent years optimizing for human psychology. Things like persuasion frameworks, review signals, and badge placement are all built to win over people.

Now, those same ideas need to work on AI agents too. So what do AI agents actually care about?

1) Keywords and their order matter: Changing a product description from “Floor Lamp for Living Room” to “Office Floor Lamp” increased selection rates by 80 points in GPT-4.1, 52 points in Gemini 2.5 Flash, and 41 points in Claude Opus 4.5.

2) Ratings and review counts pull weight: Even a 0.1 point rating increase meaningfully improves selection odds.

Volume matters too. Agents, like humans, gravitate toward the popular pick.

3) Sponsored labels work against you: AI agents are less likely to choose sponsored products. For them, it is a credibility signal, just not a positive one.

This is essentially SEO, but for machine buyers. The principles rhyme with what we already know: Relevance, trust signals, and competitive pricing, but the audience has changed.

Newer models are getting smarter fast: Failure rates when choosing between a discounted and full-price identical product dropped from 63.7% to 4.3% in Claude, and from 25.8% to 1% in GPT.

The practical move here is to identify which agents your customers use most, then test and tweak product pages accordingly, especially after major model updates.

How do you measure B2B content success?

The scoreboard is changing.

Organic search used to be the only game in town but there’s a new challenger:

For the first time, AI referral Traffic (17%) has overtaken organic traffic (16%) as the primary metric for B2B success. This is a seismic shift in how we value content.

Other key metrics include:

• Overall referral (13%): Links from other sites.
• Social Engagement and shares (13%): Viral potential and community buzz.
• Content pageviews (10%): Traditional “eyeballs” on the page.

The Crew’s Take: If you aren’t tracking how much traffic AI chatbots (like ChatGPT or Claude or Gemini) are sending your way, you’re flying blind.

The “search” era is evolving into the “answer” era.

What you can do: Check your analytics for “direct” traffic spikes or specific AI referral sources.

If it’s growing, your content is successfully “teaching” the LLMs that you are a credible authority.

Want more insights on emerging metrics, attribution shifts, and how to measure what matters now? Subscribe to Pro. It’s just $349/year for up to 10 teammates.