Doing SEO in the age of AI means working two surfaces at once: the classic organic results that still drive most search traffic, and a growing layer of AI answers — Google’s AI Overviews, plus assistants like ChatGPT, Gemini, and Perplexity — that increasingly sit between a searcher and your page. The tools changed. The fundamentals did not. This guide keeps those two facts separate, because conflating them is how people waste a year chasing tactics that never mattered.
AI Overviews launched in the US in May 2024 and expanded through 2025, generated by Gemini, appearing for a subset of queries and citing a rotating set of sources. That is the headline change most people react to. It is real, but it is not the whole picture.
What genuinely changed
Three things shifted in a way you can observe. First, an answer block now takes the top of the page for many informational queries, absorbing quick-answer clicks. Second, AI assistants outside Google have become a real discovery surface, where people ask questions conversationally and get synthesized answers with citations. Third, producing content is cheap now, which floods the web and raises the bar for standing out.
Each of these changes where attention lands and how much competition you face. None of them removes the need to be the page worth citing. The channels multiplied; the requirement to be genuinely useful got stronger, not weaker.
What stayed exactly the same
The core mechanics are stubbornly intact. Search still rewards relevance, clarity, structure, and authority. Links still matter — necessary but not sufficient, and thin content loses even with links behind it. Ranking on page one of classic organic remains the strongest observable predictor of being pulled into an AI answer at all. Being cited correlates strongly with already ranking well.
This is the reassuring part of SEO in the age of AI. If you were doing the fundamentals — answering questions clearly, structuring pages cleanly, earning authority — you are already doing most of what the AI layer rewards. There is no separate secret channel with its own rules.
Links still play their familiar role, with the familiar caveats. They remain a genuine signal of authority, but they are necessary rather than sufficient — thin content behind strong links still loses, because the content itself has to carry the value once a reader arrives. Rankings still jitter a couple of positions day to day as normal noise, so trends matter more than any single reading. And third-party volume and difficulty figures are still estimates, useful for prioritizing but never a substitute for your own data. The age of AI did not repeal any of these; it layered new surfaces on top of them.
It is worth stating why this stability exists. AI answers have to draw from somewhere, and that somewhere is the web that classic search already indexes and ranks. The retrieval that feeds an answer leans on the same relevance and authority signals search has always used. So an entirely separate optimization discipline for AI, disconnected from ranking, would be optimizing for a channel that ultimately pulls from the channel you already know. The two are not rivals; one sits on top of the other.
The honest limits of AI-answer optimization
Here is where discipline matters. The exact triggers for an AI Overview are unknown. Citation is not guaranteed and the cited source set shifts day to day. There is no real “share of voice” number for AI Overviews — any exact figure is a model’s guess dressed as data. You cannot force inclusion, and inclusion is unstable.
The measurement reality is the hard wall. Search Console folds AI Overview impressions and clicks into standard web search with no filter to isolate them, and no tool reports an exact AI Overview position. When a vendor shows you a precise number for your AI visibility rank, they are inferring, not measuring. In the age of AI, the most valuable habit is telling the difference between a metric and a guess.
AEO and GEO: adopt the sound, shelve the speculative
Answer Engine Optimization and Generative Engine Optimization are real emerging practices, but their advice ranges widely in quality. The mechanically sound part overlaps with good SEO: clear question-and-answer structure, factual accuracy, coverage of the sub-questions a topic implies, and genuine topical authority. Do these, because they help across every surface.
The speculative part deserves skepticism. Files like llms.txt, entity-stuffing, and adding schema specifically to win citations are unproven — no major answer engine has confirmed they change inclusion. Schema is still worth adding for its established benefits in helping machines understand your page; just do not treat it as a citation lever. Adopt what stands on its own merit; shelve what depends on unverified claims.
How AI changes your workflow, honestly
The biggest practical change is production speed. AI can research and draft faster than any team could by hand, which is why the web is flooding with content. That cuts both ways: you can move faster, but so can everyone, so the differentiator becomes quality control, not volume. The winning setup is AI that drafts and deterministic checks that decide what actually ships.
That is the principle behind a disciplined workflow: the AI writes, and rules decide what publishes. Validated structure, minimum depth, accurate titles and meta, and human judgment on the substance. Speed without gates just adds to the flood. Speed with gates lets you publish more of the well-structured, genuinely useful content that both classic search and AI answers reward.
The other workflow change is where you point human attention. When drafting was slow, most effort went into writing. When drafting is fast, the scarce resource becomes judgment — verifying claims, sharpening the angle, deciding whether a piece earns its place at all. Teams that thrive in the age of AI redeploy their people from typing to editing and fact-checking, because that is where the remaining value sits. The writer that never tires still cannot tell you whether it is right, and that gap is exactly where a human belongs.
A playbook you can run today
Make it concrete. Keep doing the fundamentals: answer questions directly and early, structure pages cleanly, build topical authority across clusters, and earn relevant links. Prune thin pages that only ever answered a one-liner, since those are most exposed to the answer block. Invest the freed time in depth — original testing, real numbers, clear recommendations — that a summary cannot replace.
Watch the AI layer without obsessing over unmeasurable numbers. This is where a workspace helps. SEO Rocket tracks brand mentions across ChatGPT, Google AI Overviews, Gemini, and Perplexity — mention counts with real example questions — so you can see where you show up in AI answers at all. It connects Search Console and GA4 as ground truth beside third-party estimates, and its AI writer produces validated, well-structured content of the kind these systems tend to cite. Competitor share-of-voice for AI visibility is on the roadmap, not shipped, and we would rather say so than pretend.
SEO in the age of AI rewards the same operators it always did: the ones who answer clearly, structure cleanly, build authority, and measure their own ground truth. The tools got faster and the surfaces multiplied. The job — be the page worth citing — did not change. Run the fundamentals with discipline, adopt the sound AI practices, ignore the sold certainty, and you are ahead of the field.
