“Stop watching gurus telling you what to do.”

Myriam Jessier doesn’t ease into conversations. She’s a technical SEO strategist, co-founder of Pragm, and possibly the most irreverent voice in search right now. And she’s tired of watching marketers fail because they’re following advice that worked in 2019.

“There’s best practices for SEO out there,” she tells me, “and then there’s your business reality. The stuff that actually works for you.”

Here’s what she means: while you’ve been optimizing for rankings, the entire game changed. AI didn’t just speed up search. It collapsed the distance between how machines read your brand and how humans experience it.

And most companies have no idea it’s happening.

AI Search Isn’t a Problem. It’s a Situation.

Myriam draws a distinction I can’t stop thinking about.

“Do you know what the difference is between a problem and a situation?”

A problem has a solution. It might be expensive, time-consuming, resource-intensive. But it can be solved.

A situation simply is. GDPR isn’t a problem. It’s a legal framework. You don’t solve it. You operate within it.

“Search being AI-powered is not a problem,” she says. “It is a situation.”

This reframe matters because most marketing teams are still treating AI search like something they can fight their way through with better content, more keywords, smarter technical SEO.

They’re wrong.

“The longer we operate on this premise of ‘I have lost something and I’m upset,’ the longer we are closing ourselves off to figuring this out.”

AI search is physics now. You don’t optimize for gravity. You build for it.

Thanks for reading Playbook Broken! Myriam brings the heat in this post - please share it!

The KGB Problem

Here’s Myriam’s analogy that stopped me cold:

“You as a brand should not be like the KGB bugging every single seat and trying to figure out what is going on for each person.”

Traditional SEO was about control. You controlled your content, your keywords, your metadata. Google played matchmaker, but you controlled the inputs.

That’s over.

Now you have an AI brand ambassador showing up between you and your customer, deciding what’s relevant based on context you can’t see. Chat history. User preferences. Inferred intent.

“You cannot force people to pay attention to you,” Myriam says. “You should be doing your job in a way that makes sense in the user’s context. You should be relevant in that context. The end.”

This is where most brands are getting it wrong. They’re still trying to rank first for “cancer” when what they should be asking is: In what contexts does our brand actually belong?

While You Weren’t Looking, SEO and Brand Became the Same Thing

“If you want to get really technical about it, this is technical branding.”

Myriam drops this like it’s obvious. It’s not obvious. Nobody’s talking about this.

“What you are asking your SEO to do is to understand your brand DNA, and they didn’t care about it at all for years. And now you want them to pull magic. It doesn’t work that way.”

Here’s what she means: AI doesn’t just crawl your website anymore. It synthesizes your entire brand presence—your PR pushes, your Reddit threads, your leaked memos, your CEO’s unhinged tweets—and creates a narrative.

You don’t control that narrative. But you better understand it.

Myriam’s framework has four parts:

* Your Known Brand: What you put out. Your campaigns, PR, official messaging. This is what you control.

* Your Latent Brand: What people say about you. If you have a subreddit called “F**k Nestle” with hundreds of thousands of followers, that’s latent brand. You don’t control it, but AI is absolutely ingesting it.

* Your Shadow Brand: The stuff you know internally that hasn’t come out yet. Leaked memos. Lawsuits. That thing your CEO said in a closed-door meeting. But also—your quiet DEI programs, your sustainability initiatives that aren’t press-released yet.

“Can you discreetly start feeding the LLM some information so that when ChatGPT comes up and says, ‘No, but they actually are,’ there’s data that can point to it?”

* Your AI-Narrated Brand: The synthesis of all three. What the machine actually says about you when someone asks.

“That’s your AI brand. It’s not what you say. It’s those three bits together.”

Most brands are optimizing the first quadrant and ignoring the other three. Then they’re shocked when ChatGPT describes them in ways they’d never describe themselves.

Thanks for reading Playbook Broken! Subscribe to get this in your inbox!

The Practical Move Nobody’s Making

Myriam doesn’t do theory without action. Here’s what she tells CMOs to do right now:

Test 1: The Packaging Test Take a picture of your product packaging. Ask ChatGPT what brand it is. Ask it why someone should buy it.

“If you see that it’s not answering properly, you have a problem.”

Test 2: The Logo Test Upload your logo to Google Cloud Vision API (free demo). If Google recognizes your entity and says “logo of [your company],” your technical brand is working. If it doesn’t, you’re invisible to AI.

Test 3: The Sentiment Test Take your product images with humans in them. Run them through Google Cloud Vision API or an LLM. Check the confidence scores on detected emotions.

“If you’re looking for a fun, flirty summer dress, Zara models with resting b***h face don’t scream fun.”

Test 4: The Reviews Test Take all your reviews. Feed them to an LLM. Ask for sentiment analysis.

“There’s somebody in your company that knows how to do this.”

These aren’t theoretical exercises. These are diagnostic tests to see if your brand actually exists in the way AI systems understand brands.

If you fail these tests, your SEO team can’t save you. Your brand team can’t save you. Because the gap between them just became your biggest liability.

And that gap? It’s costing you customers you’ll never know you lost. Because they’re getting a version of your brand story you didn’t write.

The Cost Per Wear Problem

Before we wrapped, Myriam told me something about how she forces herself to make hard decisions.

She grew up in a well-off family, then became “dreadfully poor” at 18. Had to rebuild. Now she runs two agencies and speaks globally, but she can’t shake the trauma of living on $2 a day for food.

So she invented a metric: cost per wear.

“I can buy the $300 dress. I just have to wear it a hundred times because my cost per wear is $3.”

Fast fashion that falls apart after ten wears? That’s $5 per wear even if you paid $50. The expensive dress is actually the rational purchase.

“Of course I should be buying the expensive one.”

Here’s why this matters for marketing:

You’re making the fast fashion decision with your marketing stack. Buying tools that feel cheap but fall apart under AI pressure. Optimizing for metrics that made sense when search was about rankings.

The expensive move—the one that feels irrational right now—is investing in technical branding. Building the systems that make your brand comprehensible to machines and humans simultaneously.

That feels like overkill until you realize: you’re going to “wear” your brand positioning every single day for the next decade.

What’s the cost per wear on your current approach?

Your Playbook Is Broken

SEO and brand were separate functions because the interfaces were separate. Humans experienced your brand. Machines crawled your site. You could optimize for both independently.

AI collapsed that separation.

Now machines experience your brand the way humans do—through synthesis, context, sentiment. And humans increasingly experience your brand the way machines do—through AI-narrated summaries, zero-click answers, contextual recommendations.

The companies that win won’t be the ones with the best SEO or the best brand team.

They’ll be the ones who realize those are the same team now.

Technical branding isn’t a nice-to-have. It’s not a 2026 priority.

It’s the discipline you should have started building six months ago.

The uncomfortable question: If Myriam ran your brand audit tomorrow, what would the packaging test reveal?

Don’t tell me you need better data. Don’t tell me you need buy-in from brand. Don’t tell me you’re going to wait and see.

Take the picture. Run the test. Find out what AI actually thinks you are.

Because your customers already did.

That’s a wrap on Season 1 of Playbook Broken (the show). More posts to come while we plan season 2… in the meantime, please do share with your network!

This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit marcsirkin.substack.com