You have heard all of it by now. "SEO is dead." "Keywords are dead." "AI changed everything, throw out the old playbook." And in the same breath, "you still need backlinks and authority." No wonder you are unsure what to keep. So let's answer the real question calmly: do SEO fundamentals still matter for AI search, or can you finally stop doing half of them?
Here is the good news. Most of your foundation still works. Some of your habits do not, and they never really did. This piece is a clean keep-or-drop audit, grounded in how AI engines actually retrieve and cite pages, not in folklore. By the end you will know exactly which fundamentals still earn citations in tools like ChatGPT, Perplexity, and Gemini, and which ones to quietly retire so you stop spending effort where it no longer pays. Because the honest answer to "does SEO matter for AI search" is not yes or no. It is "keep the foundation, drop the tricks."
The short answer, in one table you can scan
Keep the foundation. Drop the tricks. That is the whole audit in one line, and it holds up under scrutiny.
If you only skim one thing today, skim this. Here is what seo carries over to ai search, and what does not, side by side.
| Keep (still earns AI citations) | Drop (obsolete or counterproductive) |
|---|---|
| Technical crawlability and rendering | Keyword stuffing and exact-match repetition |
| E-E-A-T (experience, expertise, trust) | Keyword density as a score to chase |
| Topical authority (depth on a subject) | The meta keywords tag |
| Information gain (original data, first-hand insight) | Domain Authority as the deciding factor |
| Search intent alignment | Schema added expecting a citation boost |
| Clear structure (headings, lists, tables) | Thin pages built on keyword variants |
| External mentions and third-party citations | Link schemes and reciprocal-link networks |
| Meta titles and descriptions | Optimizing for a single AI engine |
Notice the pattern. The keep column is everything that makes a page genuinely accessible, credible, and useful. The drop column is everything that tried to trick an algorithm rather than serve a reader. AI did not kill SEO. It killed the shortcuts.
So when someone tells you the seo vs aeo fundamentals are completely different disciplines, gently push back. The fundamentals underneath are largely the same. What changed is the goal on top: from "rank for a keyword" to "become the source a machine quotes."
Why the answer lives in how AI retrieves and cites
Every keep-or-drop call makes sense once you see the pipeline. So before the audit, a two-minute tour of how an AI answer gets built. Follow this and you will never have to memorize a ranking-factor list again.
AI engines like ChatGPT, Perplexity, Gemini, Claude, and Google AI Mode pull sources through a retrieval-augmented pipeline. It runs in three stages, and each one explains a column in that table.
Stage one is retrieval. The engine turns your prompt and candidate pages into vectors, then finds the closest matches by meaning, not by matching words. This is why exact-match stuffing is dead. Repeating a phrase does not make your page more semantically relevant, and it can make your writing worse. It is also why technical accessibility is non-negotiable. If an AI crawler cannot fetch and render your page, that page never enters the pool, no matter how good it is.
Stage two is re-ranking. The engine scores the pages it retrieved for relevance and for information gain, meaning how much a passage adds beyond what everything else already says. If five pages repeat the same advice, the engine leans toward the one that brings a fresh data point or a first-hand example. This is the mechanism that turned "information gain" from a nice idea into the operative replacement for keyword optimization.
Stage three is the citation decision, and it runs on risk. The engine prefers sources that look verifiable: clear authorship, a credible domain, original data, named entities. It is trying not to be wrong. This is why E-E-A-T signals and outside mentions decide who gets cited among otherwise similar pages, and why a generic, unsourced draft rarely gets picked at all.
One number shows the shift in motion. Analyses of AI Overview citations found that a large majority once came from pages ranking in Google's top ten. A year later, that share had roughly halved. Traditional rankings clearly still help, but AI engines are increasingly surfacing sources through their own retrieval logic, rewarding relevance and gain over raw ranking position. So when you ask does seo matter for ai search, the honest read is yes as a strong foundation, and no as a guarantee. The floor moved up; the ceiling changed shape.
Hold onto that shape. Access, then relevance and gain, then trust. Those are your seo basics for aeo, and every recommendation below traces back to one of the three.
What carries over: the keep column, up close
Keep the work that makes your page fetchable, credible, and easy to lift. That is the through-line of everything here.
Start with technical SEO, because nothing else matters if this fails. Crawlability, clean rendering, a working sitemap, fast pages, and a robots file that does not accidentally block AI crawlers like GPTBot or PerplexityBot. A study of technical factors in AI search found these still play a real role, just not always the way teams expect. It is less about where a keyword sits on the page and more about whether a bot can reach and parse the content at all. If your site leans heavily on client-side JavaScript, this is worth a serious look, because content that only appears after the browser runs code can be invisible to the machines you want quoting you.
Next, E-E-A-T. Experience, expertise, authoritativeness, and trust matter more in AI citation than they did in plain ranking, because the engine is risk-averse. It wants a source it can stand behind. Named authors, real credentials, and evidence of first-hand work all raise your odds. So does topical authority, which is simply depth and breadth on a defined subject. Practitioners tracking citation data keep landing on the same finding: depth of coverage is the single strongest predictor of getting cited again and again. One deep, connected cluster beats twenty shallow posts.
Then there is information gain, and this is the one worth circling. Original research, proprietary data, a real case study, a framework nobody else has named. When your page adds something the rest of the internet cannot copy, the re-ranking stage rewards you. Consensus content gets skipped. This is also why what seo carries over to ai search includes your editorial point of view. A distinct angle is itself information gain.
One more carries over harder than ever: being mentioned and cited by other trusted sites. External validation was always an SEO signal. In AI citation it is close to decisive, because the engine treats third-party corroboration as proof you are real and worth quoting. Research suggests your own site is often only a small share of what AI cites for your brand, so reviews, press, and partner pages end up doing heavy lifting you cannot do from your own domain alone.
Two more that quietly survived. Search intent alignment still matters, but think in meaning rather than exact wording: match what the prompt is really after. And your meta titles and descriptions still earn their keep. They shape click-through and help AI systems classify your entity and topic. Keep writing them with care.
What to drop: the habits that stopped paying off
Drop anything whose only job was to game an algorithm. If a tactic never helped a reader, it almost certainly does not help a machine either.
Keyword stuffing and density targets go first. Semantic retrieval does not care how many times you repeated the phrase, and over-stuffing can actually hurt how coherent your passage reads. Stop chasing a density number. The meta keywords tag goes too. Google confirmed years ago that it does not use it for ranking, and AI engines ignore it as well. It is dead weight.
Now the harder one to let go of: Domain Authority. It is fine for a rough sense of discovery, but it is not the deciding factor for a citation. A high-DA page on an irrelevant topic gets passed over. A lower-DA page with a perfectly relevant, high-gain passage gets cited. Niche authority beats raw DA now. So do not treat your DA score as the finish line.
Schema deserves a careful word, because the advice online is loud and contradictory. Keep your structured data in place. It helps machines disambiguate entities and understand your page. But do not add JSON-LD expecting a jump in citations. A controlled experiment tracking pages that added schema found citations did not move on any platform. Treat schema as baseline hygiene, not a lever you crank for more visibility.
The rest of the drop list is the old black-hat and gray-hat toolkit, and it ages badly. Mass-producing thin pages around minor keyword variants. Publishing unedited AI drafts that add zero new information. Link schemes built purely to inflate authority signals. These violated search guidelines before, and they raise exactly the risk flags that make an AI engine skip you now. One more to add for the AI era: do not optimize for a single engine and assume the rest behave the same. Citation sets overlap far less than most marketers expect, so betting everything on one platform leaves most of the real estate untouched.
It helps to know the engines lean differently. ChatGPT and Claude tend to play it safe, favoring established, high-authority domains and third-party brand mentions, and they are slower to cite an unknown source. Perplexity is the most adventurous of the group, more willing to cite newer or smaller sources when the page delivers strong relevance and unique data. Gemini and Google AI Mode stay closest to traditional Google rankings, so your classic technical work pays off there most directly. The seo basics for aeo are the same underneath all four; the appetites on top differ. That gap is exactly why one-engine optimization quietly misleads you.
The one genuinely new fundamental: information gain
If you add one new habit this quarter, make it this: give every important page something the internet does not already have.
Information gain was a theory in SEO for years. AI made it mandatory. In plain terms, it measures how much your content adds beyond the existing consensus on a topic. The re-ranking stage rewards it directly, which is why it has become the practical successor to keyword optimization.
The good part is that you already have raw material for it. You are closer than you think. Original sources of gain that are hard for competitors to copy include:
- Original research: a small survey, an experiment, an analysis of your own numbers.
- Proprietary data: product usage patterns, aggregated customer behavior, support trends.
- First-hand case studies from real work, with specifics and outcomes.
- A clear, reasoned point of view that departs from the consensus.
- A named framework or mental model for a problem people keep hitting.
The sweet spot is consensus plus gain. Pure consensus gets ignored because it adds nothing. Pure contrarian-for-its-own-sake gets distrusted. Confirm what is true, then add the one thing only you can say. That is the move.
Do not let this feel like a mountain. You do not need a research department. You need one honest data point or one real story per key page, and you build from there.
What this means for your week
Start with access, then trust, then gain, in that order. That sequence keeps you from optimizing a page that no crawler can even reach.
Here is a calm, doable path. First, confirm AI crawlers can fetch and render your top pages, and that nothing in robots.txt is quietly blocking them. Second, pick your handful of highest-intent pages and strengthen the trust signals: real author, clear structure, crisp answers near the top of each section so a passage can be lifted cleanly. Third, add one piece of genuine information gain to each of those pages. Fourth, look beyond your own domain, because research suggests a brand's own site is often only a small slice of what AI cites for it. Third-party mentions, reviews, and press carry a lot of the weight.
That last point is where measurement comes in. You cannot improve what you cannot see, and single-platform screenshots will mislead you. If building a tracking process from scratch feels like one more thing you do not have time for, this is a place a platform earns its keep. DeepSmith's AI visibility tracking lets you define the prompts your buyers actually ask, then watch mention and citation rates across ChatGPT, Gemini, Perplexity, Claude, and Google AI Mode, and see which competitor pages are winning. It hands you the framework instead of asking you to invent one.
And once you know which pages to fix or write, the same platform can produce them with keyword coverage, clear structure, internal links, and metadata built in during creation rather than bolted on after. That is not a magic button. It is the manual work around the writing, handled, so your time goes to judgment instead of formatting.
None of this requires throwing out your SEO knowledge. The seo vs aeo fundamentals share a spine. You are repurposing what you already know: keep the foundation, retire the tricks, and add information gain on top.
The throughline
So, do seo fundamentals still matter for AI search? Yes, more than the "SEO is dead" crowd wants to admit. The foundation of technical health, authority, intent, and structure still decides who gets cited. What changed is the layer on top. Keyword stuffing, density chasing, the meta keywords tag, schema-as-a-lever, and DA-as-a-silver-bullet have all aged out. In their place: information gain, topical depth, third-party validation, clean semantic structure, and tracking across every engine that matters.
You do not need to start over. You need to keep the right things and let go of the rest. Pick one high-intent page this week, make sure a crawler can reach it, and add one thing only you can say. That is the whole game, one page at a time.
Ready to see where you show up in AI answers and close the gaps with content built to be cited? Start a free DeepSmith trial and watch your first real data come in.
