Run any site through a crawler and you’ll get a report with 300 issues, 40 warnings and a color-coded severity scale that treats a missing alt tag like a house fire. Most teams respond in one of two ways: they try to fix everything and burn a quarter doing it, or they look at the wall of red, feel defeated, and go back to publishing blog posts. Both responses cost you the same thing — the small number of technical problems that are genuinely holding the site back never get touched.
Technical SEO has an overwhelm problem, not a complexity problem. The work that matters is a short list. The rest is noise generated by tools that have to justify a subscription. After more than a decade doing SEO and online reputation work for startups, law firms and public figures, the pattern I keep seeing is that sites rarely lose because of an obscure technical flaw. They lose because something basic is broken and nobody noticed — or because the content is thin and everyone is busy arguing about render-blocking JavaScript.
Why technical fundamentals matter more in the AI era, not less
There’s a fair argument that technical SEO got less glamorous over the past decade. Search engines got better at crawling messy sites, forgiving duplicate URLs and rendering JavaScript. A lot of what used to be edge-case optimization became table stakes handled by your CMS.
Then generative search arrived and quietly raised the stakes again. Generative Engine Optimization (GEO) — the practice of making your content visible and citable inside AI answers — depends on a pipeline that starts with a machine successfully fetching and parsing your page. Google AI Overviews, ChatGPT’s browsing, Perplexity and the rest are all working from retrieved documents. An AI citation is a link in an answer, and no system cites a page it couldn’t read.
That matters because AI crawlers are, generally speaking, less patient than Googlebot. Many don’t execute JavaScript the way a modern browser does. Many won’t wait around for a slow server. Many read your raw HTML and make a fast judgment about what the page is actually about. If your content only exists after a client-side render, or your key answer is buried inside a tabbed widget, you’ve made yourself unreadable to exactly the systems that are increasingly forming the answer before anyone clicks. This is one of the more direct ways that SEO feeds AI search — the same infrastructure serves both.
The way I think about this: technical SEO is no longer about squeezing out ranking points. It’s about removing the reasons a machine might skip you.
How to triage: three tiers that decide where your time goes
In my work with clients, I sort every technical finding into one of three buckets before anyone opens a ticket. It turns a 300-item audit into a two-week plan.
Tier 1: Can it be reached and read?
These are the fixes where a failure means zero visibility — not diminished visibility, zero. Nothing else is worth doing until these are clean.
- Indexation reality check. Does the number of indexed pages roughly match the number of pages you want indexed? Large gaps in either direction point to a real problem — orphaned content, accidental noindex tags, or thousands of junk URLs diluting the site.
- Robots and meta directives. A stray disallow rule or a noindex left over from a staging environment is the single most common catastrophic issue I find. It’s also the fastest fix.
- Server-rendered content. View the page source — not the inspector, the source. If your main body copy isn’t there, you’re gambling on every crawler’s rendering budget.
- Status codes and redirects. Broken internal links and long redirect chains waste crawl budget and break the signal path between your pages.
- One canonical URL per piece of content. Trailing slashes, parameters, HTTP and HTTPS versions — pick one and enforce it.
Tier 2: Does the structure make meaning obvious?
This tier is where technical work starts overlapping with content strategy, and where most of the upside lives right now.
- Heading hierarchy that describes the page. One H1, H2s that read like the questions a person would actually ask. A language model parsing your page uses that structure to decide what the page answers.
- Schema where it’s genuinely applicable. Organization, Article, FAQ, Product, LocalBusiness. Structured data doesn’t buy you a ranking, but it removes ambiguity about entities, authorship and dates — and ambiguity is what gets you passed over.
- Internal linking as an architecture, not an afterthought. Your link graph tells crawlers which pages are important and how topics relate. I’ve written separately about internal links as a ranking and citation signal, because this is the lever most teams underuse.
- Clean, stable URLs. Descriptive, lowercase, and not changing every time marketing rebrands a section.
Tier 3: Everything else
Image compression, minor Core Web Vitals tuning, hreflang refinements on a single-market site, alt text on decorative icons. This work is legitimate. It’s also where teams hide when they don’t want to deal with the fact that their content isn’t good enough. Do it on a maintenance cadence, not as a strategy.
A concrete example of what this looks like in practice
Consider a typical SaaS site built on a modern JavaScript framework. The audit flags 200 issues. The team panics about image formats and a handful of missing meta descriptions.
Meanwhile, the actual problems are three: the documentation section renders entirely client-side, so half of the site’s most useful content is invisible in raw HTML; the blog generates a tag archive for every keyword an author ever typed, producing hundreds of near-empty pages competing with real articles; and the pricing page is linked only from a dropdown menu that doesn’t exist in the source.
Fix those three and you’ve changed what the site can do. The other 197 issues still sit in the report, and it doesn’t matter. That’s the entire discipline — knowing which findings are load-bearing.
One thing that consistently works: before you touch a single technical item, take five of your most important pages and check whether a machine reading only the HTML would understand what each page is, who wrote it, what it claims and why it’s credible. If the answer is no, that’s your roadmap.
The Cory Maki SEO approach: technical work in service of citability
What I’m seeing across AI search is that the technical layer and the credibility layer have stopped being separate projects. A page that’s fast, server-rendered and clearly structured is a page that’s easy to quote. A page that’s ambiguous about who’s speaking and when it was published is a page that gets skipped in favor of one that isn’t.
That’s the logic behind the ARC Method and the citation frameworks I work from — clarity and structure are what make content citable, and no amount of clever optimization substitutes for being legible. It’s also why technical hygiene alone won’t get you there. Clean infrastructure gets you considered. Topical depth over content volume gets you chosen. The two have to run together.
If you want a durable filter for every technical decision, use this one: does this change make it easier for a machine to find, read and correctly attribute my content? If yes, it’s worth doing. If you can’t explain how it does, it’s Tier 3 — schedule it and move on.
A practical starting sequence
- Week one: verify indexation, check robots directives, confirm your key pages exist in raw HTML, fix broken internal links on your top 20 pages.
- Week two: clean up thin auto-generated archives, enforce canonicals, add Organization and Article schema with real author data.
- Week three: map your internal link structure around topic clusters rather than publication date.
- Ongoing: monitor server response times, watch for new crawl errors monthly, and check whether AI assistants can retrieve and summarize your cornerstone pages accurately. If they can’t, ask why.
None of this is exotic. Most of it is the same list of SEO fundamentals that still matter, now carrying more weight because a second set of readers arrived. If you’re thinking about Generative Engine Optimization as a separate discipline from technical SEO, you’ll end up doing both badly. It’s one pipeline: crawl, parse, understand, cite.
The durable principle is unglamorous and it hasn’t changed in a decade of this work. Make your site easy to access, easy to parse and honest about what it is. Then spend the time you saved on content worth citing.
Photo by Vitaly Gariev on Unsplash