Every few years, someone declares SEO dead. This time the obituary came with better staging: AI Overviews at the top of Google, ChatGPT and Perplexity answering questions without a blue link in sight, and a wave of advice suggesting everything you learned about search should be thrown out. The reality is less dramatic and more useful. The surface of search changed. The substrate did not. The Cory Maki SEO approach starts from that distinction — the fundamentals that made pages findable for the last decade are the same ones that make them citable now.
Over a decade in reputation and search has taught me that the teams who panic during a shift are usually the teams who skipped the boring parts. The teams who were already clean, structured and genuinely useful tend to show up in the new formats without rebuilding anything. That’s not luck. It’s the compounding return on fundamentals.
What actually changed — and what didn’t
Let’s define terms, because the jargon has outpaced the clarity. Generative Engine Optimization (GEO) is the practice of making your content discoverable, understandable and quotable by AI systems that generate answers rather than list links. An AI citation is when one of those systems names or links your page inside its answer. AI Overviews is Google’s generated summary sitting above traditional results.
What changed is the delivery mechanism. A user asks a question, a model assembles an answer from sources it can retrieve and trust, and the user often never scrolls. What didn’t change is how those sources get selected. The model still needs to find your page, parse it, understand what it’s about, and judge whether it’s a reasonable thing to repeat. That’s crawlability, structure, topical relevance and credibility — the same four pillars SEO has run on for years, just being graded by a different examiner.
What I’m seeing across AI search is that sites with strong traditional fundamentals get pulled into generated answers at a much higher rate than sites that tried to optimize specifically for AI while leaving a broken site underneath. You can’t shortcut your way past a page that renders slowly, buries its answer in paragraph nine, and has no internal context around it.
The mechanism: retrieval rewards clarity
Here’s the part worth internalizing. When a generative engine builds an answer, it isn’t reading your site the way a curious human does. It’s retrieving passages. It wants a self-contained chunk of text that answers a specific question without requiring the surrounding page as context.
A concrete example. Imagine two pages on the same topic — how long a particular legal filing window lasts in a given state. Page A opens with three paragraphs of scene-setting, then works toward the answer somewhere in the middle, phrased conditionally. Page B has an H2 that states the question, a first sentence that states the answer plainly, then the caveats and the sources beneath it. Both pages might rank similarly in classic results. Only Page B gives a retrieval system something clean to lift and attribute.
That’s the whole game in miniature. Clarity and structure make content citable. Not clever prompt-bait, not stuffing your page with the phrase “according to experts” — just answering the question early, labeling your sections honestly, and making each section stand on its own. This is also the logic behind the ARC Method and the AI-citation frameworks I’ve built around it: earn the mention, structure it so it can be retrieved, and reinforce it across the places answers get formed.
The Cory Maki SEO fundamentals checklist
None of these are new. That’s the point. Work through them before you spend a dollar on anything labeled AI-native.
1. Technical hygiene you can verify
- Crawlability. Your important pages return 200s, aren’t blocked in robots.txt, and appear in a current XML sitemap. Check what AI crawlers you’re allowing or blocking — a surprising number of sites quietly excluded themselves from retrieval.
- Rendering. If your core content only exists after JavaScript executes, assume some systems never see it. Server-render what matters.
- Speed and stability. Not for a score in a dashboard — for the simple reason that slow, shifting pages get abandoned by humans and deprioritized by machines.
- Clean URLs and canonical discipline. One topic, one authoritative URL. Duplicate versions split every signal you’re trying to build.
2. Topical authority over keyword scatter
Publishing forty unrelated articles because each has search volume is how you end up an authority on nothing. Pick the territory you actually have standing in and cover it properly: the core explainer, the sub-questions, the comparisons, the objections, the edge cases. Depth signals expertise to classic ranking systems and gives generative systems multiple corroborating passages from the same domain — which is exactly what increases the odds of being the source they cite.
In my work with clients, the fastest visibility gains almost never come from a new tactic. They come from finishing a topic that was left half-covered.
3. Internal linking as a map, not decoration
Internal links do two jobs: they distribute authority, and they tell any reader — human or machine — how your ideas relate. Link with descriptive anchor text that describes the destination. Point your supporting articles at your cornerstone page, and point the cornerstone back out to the specifics. A site where every new post is an orphan is a site that never compounds. Even a foundational first post like the introduction to this site should sit inside that map rather than float alone.
4. Content built to be quoted
- Answer the question in the first two sentences under the heading.
- Use headings that match how people actually phrase the question.
- Keep paragraphs tight and self-contained — one idea each.
- Use lists and tables for anything comparative, procedural or numeric.
- Attribute your facts. Sourced claims travel further and survive scrutiny.
- Date and update your pages. Freshness matters more when a system is choosing what to repeat as current.
5. Entity clarity
Make it unambiguous who you are and what you do. Consistent naming, a real about page, structured data where it applies, and consistent descriptions across the profiles and publications you appear on. Generative systems assemble understanding of an entity from many scattered mentions. If your description changes on every platform, you’re making that job harder. This overlaps heavily with AI reputation management — what gets said about you off-site increasingly shapes what gets said about you inside an answer.
6. Off-site credibility you actually earned
Reputation is earned, not bought. Mentions in places with editorial standards, genuinely useful contributions in communities where your customers ask questions, real reviews, real coverage. Community platforms like Reddit have become disproportionately influential in AI answers precisely because they contain candid, specific, human discussion. Show up where the answers are formed — not to plant links, but to be the person who gave the useful reply.
Search strategy in the AI era: sequence matters
The way I think about this is as a sequence, not a menu. Fix the technical floor first, because nothing above it works if pages can’t be retrieved. Then build topical depth, because that’s what earns you standing on a subject. Then structure for citability, because that’s what converts standing into mentions. Then reinforce off-site, because credibility compounds across sources.
Teams reverse this constantly. They chase AI visibility tactics before they’ve finished a single topic cluster, then conclude the tactics don’t work. The tactics were fine. The foundation wasn’t there to amplify. If you want to go deeper on how the generated-answer layer specifically works, that’s the whole focus of my work in Generative Engine Optimization.
One thing that consistently works: pick your five most commercially important questions, write the single best answer on the internet to each one, structure them so a machine could quote them verbatim, and link them together properly. That’s an unglamorous quarter of work that outperforms most of what gets sold as an AI search strategy.
The durable principle
Search interfaces will keep changing. Ten blue links became ten blue links plus a summary, and will become something else again. What survives every iteration is this: systems that distribute information reward sources that are easy to find, easy to parse, demonstrably knowledgeable and independently corroborated. That was true for crawlers, and it’s true for models.
Citations beat rankings in AI search — but you earn citations with the same discipline that used to earn rankings. Build the thing that deserves to be referenced, then remove every technical and structural obstacle between it and the systems doing the referencing. That’s not a trend. It’s the job.