SEO and Generative AI: What Actually Changes

Generative AI is changing how content is discovered and presented, not the qualities that make that content worth trusting. Generative engines like ChatGPT or Google's AI overviews largely rely on content that's already indexed, structured and factually verifiable to build their answers. A site that's invisible on Google stays invisible to these tools too.

The problem: mistaking a new technology for a methodological break

Since generative search engines arrived, an idea has often circulated in exaggerated form: that classic SEO is dead, replaced by an entirely new discipline requiring unheard-of techniques. That idea deserves nuance. What genuinely changes is the discovery surface (a generated answer rather than a list of blue links); what doesn't change is the need for content that's clear, accessible to crawlers, and reliable enough to be picked up.

It's also worth being honest about the limits of what can be claimed here: nobody outside the teams at OpenAI, Anthropic or Google knows precisely the internal mechanisms that determine which source gets cited and why. The recommendations below rest on observed, reasonable principles, not on a guaranteed formula.

What doesn't change

Indexing remains the entry point

A generative engine that draws on web search (which is the case for most everyday uses) can only cite content it can actually find. A site with indexing problems, poor technical structure, or one that's new and barely visible starts at the same disadvantage against a generative AI as it does against classic Google. See our article on why a site doesn't show up on Google.

Authority and trustworthiness still matter

Content that's consistent with what's said elsewhere on a topic, coming from an identifiable source, is still favored over isolated or contradictory content. Trust isn't built any differently than before: consistency, verifiable sources, and the absence of unverifiable claims.

Clear structure remains an advantage

Explicit headings, an answer that appears early in the text, readable lists and tables: these qualities already made life easier for a rushed human reader and for classic indexing crawlers. They also make it easier for a generative system to extract information, since it needs to quickly identify the part of the text that answers a specific question.

What concretely changes

The direct answer becomes even more decisive

A generative engine builds its answer from fragments of text. Content that answers the question within the first few sentences, in a way that's self-contained and understandable out of context, has a better chance of being picked up than content that builds its argument gradually across several paragraphs before reaching the useful fact.

The click is no longer the only goal

In classic search, the end goal is almost always the click through to the site. With a generated overview (Google AI Overviews) or a conversational answer (ChatGPT, Copilot), the user can get their answer without ever visiting the source site, while still seeing its name cited. Being cited as a source becomes a goal in its own right, alongside direct traffic, particularly for brand visibility.

Structured data becomes more important

Structured data markup (schema.org: FAQ, article, product, organization) helps automated systems, human or generative, understand precisely what a piece of content actually is. That wasn't new for Google, but its usefulness grows as more systems rely on an automated understanding of content rather than simple keyword matching.

Update frequency becomes more visible

Dated content that's never refreshed loses relevance faster in an environment where generative answers can compare several recent sources. Keeping content up to date, with verified information, becomes a more visible trust signal than before.

A concrete example of the transition

Picture a local plumbing company that has spent years publishing classic keyword-stuffed content ("best plumber Lyon", repeated several times per page). That kind of content, already ineffective in modern classic SEO, is even less effective in a generative environment: it doesn't answer any specific question, contains no usable verifiable fact, and offers nothing a system could usefully cite.

By contrast, a page that clearly answers "how much does drain unblocking cost in Lyon", with a realistic price range, the factors that make it vary, and an FAQ covering common cases, produces content that's useful for a rushed reader, well positioned for Google, and structured in a way a generative system can use. The transition therefore doesn't call for a brand-new strategy, but for reinforcing content practices that have already been recommended in SEO for years: answer precisely, with facts, rather than filling a page with repeated keywords.

What remains uncertain, and shouldn't be overstated

• There's no guaranteed method for being systematically cited by a given generative engine.
• Citation algorithms keep evolving and aren't documented publicly in any detail.
• No agency or tool can legitimately promise a specific position or citation frequency in ChatGPT, Claude or an AI overview.

Staying factual about these limits protects against unrealistic expectations, in either direction.

What to remember

• Classic SEO (indexing, structure, authority) remains the foundation that generative engines also rely on.
• Content that directly answers a question from its opening lines is better positioned to be picked up by a generative AI.
• Being cited without a click becomes a visibility goal in its own right, alongside direct traffic.
• Structured data and regular content updates gain importance in this environment.
• Nobody knows precisely how generative engines' citation algorithm works: it's better to rely on solid principles than on unverifiable promises.

Frequently asked questions

Has traditional SEO become useless with generative AI? No. Generative engines largely rely on the same signals as classic search (indexed content, clear structure, domain authority) to decide what to cite. A site that's poorly indexed or invisible on Google has just as little chance of being cited by a generative AI.

Do we know the exact algorithm that determines what ChatGPT or Claude cite? No, and nobody outside those companies knows it precisely. The available observations point to general tendencies (clear, structured, factual content) rather than a guaranteed formula, despite some commercial promises you may read on the subject.

Should content be created specifically for AI rather than for Google? Not really as a separate track: content that clearly answers a question from the start, with a readable structure and verifiable facts, works just as well for a human reader, for Google, and for a generative engine. These are the same underlying qualities, not two opposing strategies.

Do Google's AI Overviews reduce traffic to websites? That's an observed trend for certain informational searches, where the user gets an answer directly without clicking through. It reinforces the value of being cited as a source within those overviews rather than relying solely on the classic click.

In summary

Generative AI shifts part of the attention from the click to the citation, but it doesn't replace the fundamentals: clean indexing, clear and structured content, a reliable source. A site built on those foundations stays well positioned, whether search happens through a classic link or a generated answer. Websites built by VeryAppi bake this content structure in from the design stage.