Author name: Dana Billingsley

The AI industry is starting to standardize how it tests itself. Brand measurement isn’t there yet, and it may not need to be.

A number without context isn’t necessarily wrong. It’s incomplete. Here’s what to ask before you trust one in a boardroom.

A visibility score tells you something moved. It doesn’t tell you why. Here’s how to find the actual cause before you try to fix it.

Three AI labs that rarely agree on anything have been meeting quietly since July. Here’s what their talks actually cover, and what they don’t.

Visibility tells you whether a brand entered an AI answer. It does not tell you whether the brand would survive the moment an AI system actually decides what to do.

Checkout inside AI Mode is the visible example, but the underlying shift, AI acting on recommendations rather than just stating them, extends well past retail.

AI can mention a brand with incomplete information. Acting on that brand, completing a cart or a checkout, appears to require a different, stricter kind of proof.

AI systems can now complete a purchase in specific, limited cases. Here’s exactly what’s confirmed, what’s still rolling out, and why it matters for how AI recommendations get measured.

A metric that can only emit a number will emit one for a real result, for a category error, and for a failed API call. All three arrive as the same red panel.

A third of what AI engines cite is not the large publications everyone is competing over. It is smaller sites nobody is bidding on.