Claude Is Already Writing Code at Anthropic — Your Brand Can't Even Control What AI Says About You

Anthropic dropped a number yesterday that's worth paying attention to.

As of August 2026, Claude now "leads" 26% of the company's AI R&D work. Not assists — leads. Over 90% of tasks reach "AI collaboration" level or above.

Meanwhile, Google just put an AI shopping agent inside YouTube. Users watching a video can now have AI pick out products, compare prices, and complete purchases for them. Lululemon and Coach are already plugged in.

And here's another one: SpaceX is discussing buying customer data from bankrupt startups specifically to train Grok. Google did something similar after Spirit Airlines shut down — it offered to buy the airline's business data for AI training.

Three stories, one reality: AI is no longer a "tool." It's becoming a business participant — running R&D, making purchasing decisions, and feeding itself data to grow.

The Reality for Brands Has Changed

Brand marketing used to come down to two things: get people to see you, get people to trust you.

Now there's a third: get AI to see you, understand you, and recommend you.

Why? Because users have changed.

More and more people don't search anymore. They ask AI. "Find me a reliable supplier for X" "Which brand offers the best value for X" "Is X better than Y" — whatever AI answers, they trust.

Google's data is clear: during testing, lululemon submitted a set of structured product information, and AI included them in 50% of relevant recommendations.

Fifty percent.

What does that tell us? AI doesn't recommend randomly. It looks at whether your information is clear enough, structured enough, authoritative enough. Give it what it needs, and it cites you. Give it a mess, and it skips you.

Most Brands' Digital Assets Are Still Built for Human Eyes

Nice websites. Polished product pages. SEO keywords stuffed everywhere.

But AI struggles to extract value from all that.

What AI needs is different: structured product specs, verifiable case studies, traceable certifications, technical documentation so clear that AI can read it without "guessing."

Think about it: when AI is answering a user's question, it needs to quickly pull together an answer from dozens of sources. If your brand's information is scattered across flashy pages that AI can't parse — you simply don't exist.

Not because your product is bad. Because AI can't "see" you.

That's Exactly What GEO Does

A lot of people think GEO is about "making AI like you." Not quite.

GEO is about organizing your brand information into a format AI can directly use.

It's not about flattering AI, gaming the system, or using some black magic. It's about presenting your core assets — product info, technical capabilities, industry case studies, certifications — in a structured way that AI can find and verify instantly.

Sounds simple, right? But 99% of brands haven't done it.

Because for the past decade, every brand optimized for search engines. Keywords, backlinks, rankings — that whole system was built for Google's crawlers. Now the game has changed. AI doesn't need you to stuff keywords. It needs clear entities, structured data, and verifiable facts.

Brands that do GEO will own their position in AI's answers.

Those that don't are essentially handing over the definition of "who they are" to AI — letting an AI that may have never properly read your website describe your brand.

The Window Won't Stay Open Long

Anthropic's Claude is already doing R&D work. Google's AI is already completing purchases for users. SpaceX and Google are buying up data to feed their models.

This isn't the future. This is today.

Every day that passes, AI's role in business decisions grows a little more. Every day, more users get comfortable "asking AI" instead of "searching themselves."

Brands doing GEO now are claiming position before the AI answer ecosystem fully solidifies.

Wait until everyone else catches on, and you're no longer claiming position — you're chasing it.

Data sources: Anthropic AI Development Metrics (September 17, 2026), Google Official Blog (September 16, 2026), Bloomberg (September 18, 2026)