Optimizing for Generative AI (GEO)

Optimizing for generative AI (also called GEO or AEO depending on who's writing) means positioning your content so AI search engines cite you as a reliable source. It's still SEO. Google made that explicit in its 2026 guidance: AI Overviews and AI Mode run on the same index, the same quality signals, and the same ranking systems as classic Search.

What that means for you: no parallel discipline to invent, no special markup to layer on, no rewriting the site "for the AIs." The work is putting the right priorities in front and stripping out the noise that a lot of agencies are still selling.

I work with US and Canadian businesses. The work is done remotely, but understanding your local market still matters for positioning your brand correctly.

01

How AI search actually works

Google AI Overview, AI Mode, ChatGPT Search, Perplexity, Gemini. These tools don't rank pages in the classic sense. They query the search index, retrieve the relevant pages (the technique is called retrieval-augmented generation), then generate a single answer while citing the most useful sources.

The bar is no longer just "are you ranking well" but "are you useful, original, and trustworthy enough to be cited in the answer."

The pages that get cited share a clear profile:

They're indexable, crawlable, and technically healthy
They offer non-commodity content: a real viewpoint, lived experience, original data, something an AI couldn't fabricate on its own
They're signed by a credible, identifiable author
They're well-organized for human readers, with clear headings and a natural structure
The entity behind them (brand, person, organization) is clearly recognizable to search systems

No magic. AI search rests on the same foundation as Google Search. What changes is the final shape of the answer presented to the user, and the fact that the right citation now matters more than the ten good rankings sitting underneath it.

02

What I actually do on engagements

My approach comes down to four pillars plus a measurement layer.

Solid SEO foundations. Crawlability, indexation, sitemap, readable HTML, performance, duplicate content cleanup, backlink hygiene, fixing the critical errors. This is the backbone, and it's what Google puts first in its official guidance on generative AI search.
Non-commodity content. The real center of gravity. Case studies with actual numbers, sharp viewpoints, lived experience told by a named author, proprietary data, content nobody else can produce. This is what Google rewards, and it's exactly what a generative AI can't duplicate by summarizing the consensus.
Clear, verifiable entity. Complete Google Business Profile, Wikidata presence, consistent LinkedIn profiles, Person and Organization schema, disambiguation from anyone with the same name. AI search needs to identify your business without effort.
Legitimate off-site presence. Where your audience actually asks its questions (Reddit, Quora, LinkedIn, podcasts, conferences), you contribute usefully with your real expertise and sign your name. The opposite of bulk pitching to land in listicles, which Google explicitly calls "inauthentic mentions" in its documentation.

Measurement. Monthly tracking of target prompts: am I cited, by which engine, on what, how often. Tools like Profound, Otterly, or SEMRush depending on the case, backed by manual checks. Without tracking, you're flying blind.

Recent example: for a Montreal-based SaaS client with US customers, we rebuilt 18 key pages around quantified case studies and credited authors, and earned first citations in Perplexity and ChatGPT within 8 weeks. Share of voice in AI answers for target queries went from 0% to roughly 22%.

03

Why this matters now

In 2026, AI answers aren't an experiment anymore. Google AI Overview shows up on a large share of queries. ChatGPT Search and Perplexity pick up ground every month. When an answer is generated, a handful of sources get cited. The rest go invisible, even if they would have shown up in classic results.

What's changing isn't the nature of SEO. It's the cost of inaction. A site that would have been on page 1 two years ago can easily go uncited in AI answers, simply because its content is too generic or its entity isn't clear. That's what we fix.

And to be clear: good classic SEO (technical health, topical authority, E-E-A-T, useful content) is the foundation of all of this. Not a separate discipline to bolt on. A discipline to execute well.

Optimizing for generative AI has become a market where a lot of agencies are selling myths: llms.txt as a must-have, "chunking" content, special phrasing "LLMs will trust," bulk pitching for mentions. Google publicly took that apart in its 2026 guidance.

My edge: 20+ years of SEO and marketing analytics, plus the habit of reading the official sources before recommending anything. When I propose an action, I can point to the passage in Google's documentation that supports it. When I tell you "we're not doing that," it's documented too.

I test techniques on my own properties before I bill for them. Before I recommend something to you, I've seen what works and what doesn't on real cases.

That combination of proven SEO foundations, critical reading of official sources, and hands-on testing is what produces results.

Source-grounded approach

Recommendations aligned with Google Search's official documentation, not agency folklore.

Real measurement

Tracking your presence in ChatGPT, Perplexity, and Google AI Overview, month over month.

Solid SEO foundations

20+ years of technical SEO and marketing analytics. No shortcuts, no buzzwords.

Will optimizing for generative AI replace SEO?

No. Google's 2026 guidance is explicit: optimizing for AI search is still SEO. AI Overviews and AI Mode run on the same index and quality signals as classic Search. A site with clean technical health, strong topical authority, and non-commodity content is already well-positioned. See my SEO page for the foundation.

How do I know if my business shows up in AI answers?

I set up monitoring with specialized tools (Profound, Otterly, SEMRush) that track your brand's presence in answers from ChatGPT, Perplexity, and Google AI Overview, backed up by regular manual checks on your target queries.

My site is small. Is this relevant for me?

Often yes. AI engines don't favor big sites by default. They favor reliable, original content, signed by a credible author, that answers specific questions precisely. A small site with real expertise can get cited before a much larger competitor.

Do I need an llms.txt file?

Google confirmed in 2026 that it's not required to appear in AI answers. No other major engine has publicly confirmed using it either. It's a low-cost experiment, but it's not a quick win and I won't sell it as one. Be skeptical of consultants who put it at the top of their action plan.

Should I break content into small "AI-citable" chunks?

No. Google explicitly debunked this practice (called "chunking"). AI search systems understand the nuance of normally written text. We structure content for human readers. That's it.

Does this work for a US English-speaking market?

Yes. The same principles apply in English as in French, but the content has to be written for the actual audience, not mechanically translated. My experience with US clients lets me tune strategy and phrasing to how American users actually search.

Optimizing for generative AI: get cited when AI answers questions about your industry