The semantic grid as an SEO strategy in the face of AI

I'd like to start this edition with three thoughts:

We're seeing more and more "strategies to appear on LLMs", except that for now it's pretty vague and it represents barely 2% of traffic. I know FOMO is huge, but be careful: more than ever, it's important to sort good information from bad so you don't make mistakes, and so to choose carefully who you listen to and who you read.

The art of timing: in business as in SEO, momentum is often decisive, and can make your site take off or kill it. Making the right move at the right time is also what your client pays you for.

And in fact, why rush into pseudo LLM strategies when most SEO strategies already work to rank in AIs? (I'm talking about real strategies, not nasty black-hat stuff.)

I keep seeing that you ab-so-lute-ly have to automate your SEO.

I do the exact opposite: manual keyword identification, outline, Hn structure, etc. Why? Because automation lets you produce average content. Today, what you need is to become the SERP's originality.

⚠️ So watch out if someone sells you 100% automated SEO: it's not possible on competitive SERPs. Not yet, at least.

On to today's topic: what SEO strategy should you adopt in the face of AI?

What you need to understand is that AI came down hard on those who wanted to cheat.

We ended up with the best-optimised results at the top, not the best results overall.

That is what partly put users off, not the arrival of generative AI, but the willingness of agencies and consultants to play against Google, and not enough with it.

What's happening today is an opportunity.

SEO is going to be used to bring the best content to the top, taking into account sites' topical expertise.

At least we hope so, because that was already Google's promise back then.

It was supposed to be the perfect combination: having the most interesting content at the top.

Optimisation only comes in to make the content visible.

So we realise that SEO work will no longer be only about creating content, but also about the ability to distribute that content.

A website, yes, but more than that.

I'm thinking more and more about including newsletter writing for my clients in my SEO services, sent both by email and through a platform like Substack.

Why? Because it brings clicks from outside Google, and natural backlinks through an authoritative domain name.

But to avoid falling back into past pitfalls, it's obvious that you have to PRODUCE QUALITY.

The goal is NOT to generate clicks from Google or get a backlink, but to offer our users different formats AND support our SEO. It's not (any longer) the other way round.

Agencies and consultants have to change their mindset and see Google as an ally, no longer as an enemy.

I know it always sounds a bit cooler to say you're trying to "hack the algo", but there's a good chance you won't manage to hack anything at all and you'll disappear from the SERPs.

The semantic grid

For a long time we thought SEO = dominating a whole topic.

Today, you have every interest in adopting a different strategy. The grid isn't only for Google: it's also the only way not to be part of the 88% of SEO sites that are invisible in AIs.

I've been talking for a while about the end of traffic as an SEO metric (see my current view on SEO in the face of LLMs). We're right in the middle of it.

The goal isn't to generate as much traffic as possible, but leads (I'm speaking in particular about B2B niches).

Step 1: identify your business keywords

If we assume AI will generate most of the informational content, a fundamental change that I covered in the bet on expert content in the face of the AI tide, you need to:

• Identify specific long-tail business keywords
• Think of your informational keywords as middle and bottom of the funnel for purchase intent (comparison, client case, etc.)

To do that, we choose one or two main keywords.

We don't look at what other players in the market are doing to find them, but focus on what our product/service answers perfectly.

These two keywords must fit your brand.

Before, to choose a keyword:

• volume
• difficulty

But today:

• Volume is the least relevant criterion.
• Difficulty is an indicator that works one time in three.

To choose your business keywords, I suggest several criteria to weigh:

• A direct business interest for your activity
• Must fall within the query field of your ICP
• Lets you strengthen your semantics
• An interesting CPC (if possible)
• A suitable, winnable SERP (⚠️ if there are big players, institutional ones in particular)
• Requires expertise at the writing stage (to avoid competing with AI)

Two remarks:

• You no longer need a list of 19,388,394 keywords that you'll never work on but that look nice in an Excel file.
• Do a manual search, choosing each keyword (between 30 and 50): use Keyword Planner. To go further, my full method for finding relevant keywords for LLMs details the workflow step by step.

Step 2: start from the need, not the keyword

SEO 2020: Find a keyword → create content

SEO 2026: Identify the client's need → find the right keyword → understand the intent → create suitable content

Always stay bottom of the funnel and produce expert content.

Here are 10 useful content formats:

1. Split an FAQ across several pages
2. Content curation (news, stats, infographics)
3. Merge old articles that don't perform
4. Bring in experts on the topic
5. Use reviews from other sites to build rankings
6. Create a niche tool / simulator
7. Create comparisons
8. Optimise for a local query
9. List problems + solutions (and create a template)
10. Identify the semantic entities close to your main keyword

To automate part of this semantic analysis, I rely on my proprietary SEO prompts, which I've been testing and refining for two years.

The semantic grid strategy consists of becoming the ultra-specialised expert in a precise niche, rather than trying to cover a whole topic superficially.