If you only check one highlight this week:

• B2B Marketer? How will you evolve your website when it gets disintermediated?

• Everyone else: how Claude can do analysis and create memos, slides, and spreadsheets.

-François

We need to rethink our websites: from purpose, to structure, to measurement

In the last two weeks, I:

1. spent more time using Gemini in Chrome (see next section for more).

2. played with the new Expedia and Zillow apps inside of ChatGPT (see how you can use Figma, Expedia, Coursera, Zillow in ChatGPT) - using Expedia feels like having an EA that gives you the right travel options in seconds

3. heard more marketers report a rapid decline in organic traffic, while a growing number of high-intent leads say they discovered their brands through AI

With all this, I am now convinced: we need to quickly discuss and rethink our websites’ roles and structures.

Why?

Increasingly, it’s not just initial discovery that takes place through influencers and AI chatbots; it’s also research and selection; i.e. influencers and AI also assist buyers much further down the funnel.

So when these savvy prospects hit our sites, they’re already well-informed about:

• our products/services + their strengths and weaknesses

• our alternatives and how we compare to them

• who uses them and how

• and maybe even how to negotiate

And with 800m active weekly users of ChatGPT and 450m active monthly for Gemini, marketers and developers are frenetically rushing to be featured in AI answers or capabilities, in the best light:

• Marketers are creating LLM-optimized content

• Devs are building MCP servers, and very soon, more apps for ChatGPT

We all want to reach that massive and super-engaged audience before another brand or service beats us to these questions answers.

And by doing so, we make the AI beast even stronger.

So, what's the role of our websites with all that?

Some argue that we should focus them on the last mile, meaning the bottom-of-funnel content, to finish convincing a prospect. Ideally, with a multi-modal agent that can answer their final questions.

They are right. But there is much more to consider.

Yes, we want to “seal the deal” with ROI calculators, case studies, video demos, FAQs, whatever bottom-of-funnel content is right. No doubt.

We also need our agent(s) to answer prospects’ questions instead of letting Gemini in Chrome handle that (see screenshot below, where I preferred reading the summary from Gemini vs the full site copy). We want to collect the questions and data directly and control the answers. Let’s not let another search engine or AI disintermediate us, if we can.

But our websites are also a critical expression of our brand. One we fully control, at a moment where we seem to lose more and more control.

They are where all our content that feeds these LLMs also resides, as well as the content that is designed for humans (e.g. videos).

Plus, there’s still a portion of prospects that doesn't rely heavily on LLMs or influencers for research (shrinking fast).

I don't have the full answers yet.

But these changes are happening super fast. So I welcome your thoughts and experiments, and will continue to cover that.

Try these out to form your opinion:

• Get the latest version of Chrome and hit the Gemini button at the top right of the chrome (or use the shortcut: ⌃G)

• Invoke @expedia in a ChatGPT thread (web version) and ask it to search for travel for you

My selection of tips, news and workflows

✨ Gemini in Google Chrome

Want a quick summary of an article? Don't want to peruse a full website? Want to have a voice discussion with Gemini about a page you're on?

Using Gemini in the latest version of Chrome has been such a time saver for me.

🤖 Claude 4.5 Sonnet is great at office tasks. From analysis to creating docs

Impressive capabilities showcased in this video, where a Claude user shares a question, some data and requests an analysis. After doing a deep research on pizza joints benchmarks, Claude creates:

• XLS: a financial acquisition analysis

• DOC: investment recommendation memo

• PPT: exec summary slide

Code execution and file creation are now available to all paid plans.

🔍 A great case study on AEO

Kyle Poyar shared how Docebo aces and measures their AEO efforts.

“Here are four practical takeaways to improve your AI search visibility:

1. Measure your AI search performance through self-reported attribution, branded search traffic, and share-of-voice in AI queries.

2. Late-funnel SEO is the foundation for great answer engine optimization (AEO).

3. The dirty work of AEO has to get done: updating, formatting, and structuring content for LLMs.

4. You can see tangible AEO results with a modern tech stack.”

📹 Inspiration: what one can produce with Sora 2

Impressive AI-generated video. It takes a lot of work and talent to do this. But it takes a fraction of the time and cost it would take if one were to shoot this for real.

Do you feel any anxiety when watching? I did.

🎙️ Recipes of Granola

Like most people today, you probably use an AI note-taking tool. My favorite remains Granola. It just got a very useful upgrade.

Go to a meeting note - or better, a whole folder with meetings inside - and use the / command, and select an existing recipe (and your favorite model): ready-to-use prompts + context.

My favorites so far are:

• Red Team. Really enlightening to identify potential blind spots.

• Who owes me what

• List recent to-dos

🛠️ How AI is changing DevRel

Andrew Baker, former VP of Developer Network at Twilio, gave an insightful lightning talk about how developer marketing is evolving in this AI era. Slides - Recording.

tl;dr:

Old Game:

• Ground game (e.g. hackathons)

• Online problem-solving-focused content (e.g. tutorials).

• Conventional brand building

New Tactics:

• LLM SEO: Optimize GitHub code and Reddit for AI discovery.

• Direct Channels: Maintain proprietary communication for updates.

• MCP Servers/AI Onboarding: AI agents onboard users

• LLM Docs/CLI: Simple products need LLM-specific documentation or CLI tools.

🛠️ Vibe-coding just got a lot better for me

I've been vibe-coding an app for a while. Recently, I was stuck in long loops and debugging sessions. The AI was breaking my app, and occasionally failing to deploy.

Two things changed recently:

1. I started using Augment Code’s “Enhanced prompt” feature, which turns my quick dictated prompts into detailed specs that I review before sending the coding agent to the races (you can do the same in your favorite AI if you give and keep the context in a Project)

2. Claude rolled out Sonnet 4.5. For the type of app I'm building, it's significantly faster and more accurate. I barely have to debug anymore.

What an improvement. Vibe-coding went from something that felt like taking shots in the dark to working with a much smarter AI developer. Fewer bugs, fewer loops.

The app is almost complete, stay tuned :-)

💡 Are we underestimating how quickly AI is improving?

Julian Schrittwieser argues that the exponential progress in AI is being underestimated by the general public, media, and even some experts. He compares this to the early days of the COVID-19 pandemic, when people failed to grasp the implications of exponential growth.

To support his argument, he references two key studies:

• METR's study that shows that the length of software engineering tasks that models can autonomously perform is doubling every seven months.

• OpenAI's GDPval: This evaluation measures model performance across 44 different occupations and shows that the latest models, like Claude Opus 4.1, are already performing at or near human expert levels.

He concludes that if these trends continue, 2026 will be a pivotal year, and by the end of 2027, they will frequently outperform human experts in many fields.

Final Words

Thanks for sharing these highlights with busy marketing execs around you.🙏 

Someone forwarded you this email? You can subscribe here.

François | LinkedIn 

I'm a CMO, advisor, and "CMO Wingman". Yes, that's a thing :-). Ask my clients: in this AI era, CMOs need a strategic proactive advisor more than ever. I’m former CMO at Twilio, Augment Code, Apollo GraphQL, Decibel, Udacity and Head of Marketing for LinkedIn Talent Solutions.