AI’s Real Impact on SEO is Probably Not What You've Been Told

If AI is “changing SEO,” what exactly is it changing?

That question sounds basic, but it’s where most teams skip the work. They jump straight to the loudest version of the story: rankings are dead, content is infinite, and the only skill that matters is prompting. Those takes are usually optimized for attention, not outcomes. The quieter truth is less dramatic and more annoying: AI makes the classic SEO answer “it depends” more true, because it increases variance. Great systems get better. Messy systems get messier, faster.

Making sense of all of this, is why this newsletter exist: To cut through the noise and find the signal. To explore that sweet spot where AI and SEO intersect and make teams faster and SEO as a discipline more efficient and powerful.

That’s the real shift. AI doesn’t replace fundamentals. It stress-tests them.

A common assumption floating around is that “AI search” means technical SEO matters less. As if retrieval and summaries somehow dissolve crawlability, rendering, internal linking, and indexation. But the engine still has to access your pages, process them, and decide which ones deserve to stick around.

Google’s own documentation has been consistent on this point for years: if your site is large, crawl management and index hygiene become practical constraints, not theoretical ones. Crawl budget isn’t a thing you need to obsess over on every site, but on big, frequently changing sites it becomes a real limiter. And if you’re running a JavaScript-heavy stack, Google is explicit that how it processes and renders JavaScript matters for what gets indexed and how reliably it’s understood.

Now add AI to the mix.

AI makes it easier to create pages. It also makes it easier to create the wrong pages. Your URL count can explode without anyone feeling like they “launched” something major. Programmatic templates plus AI copy can multiply thin variants, near-duplicates, and doorway-adjacent clusters at a speed your governance was never designed to handle. If you already had parameter chaos or faceted navigation spirals, AI doesn’t solve them. It gives them steroids.

This is where teams get surprised: the site still looks fine to humans. Pages render. Content reads smoothly. QA checks the top templates and gives a thumbs up. And yet, at scale, the system starts failing in places humans don’t naturally look. Canonicals drift. Internal linking intent gets muddied by “helpful” rewrites. Duplicate clusters balloon. Crawl paths become noisy. Index bloat creeps in, and then you’re in the position of asking Google to choose the best page from a pile you didn’t mean to create.

If you want a non-theoretical example of what “spiral” looks like, faceted navigation is the evergreen one. Even when implemented well, it requires explicit controls to avoid crawl inefficiency and index bloat, controls that become more critical when you’re also generating content variants at scale.

To be fair, AI can help on the technical side. It’s good at clustering patterns in logs, flagging template anomalies, and mapping internal linking gaps when you give it real inputs. But it’s also good at making it easier to ship changes without understanding system behavior. (more about this next week) The danger isn’t that AI is “bad at SEO.” The danger is that it reduces friction in the places where friction used to protect you.

The easiest mistake to make with AI content is to confuse speed with advantage.

Yes, AI makes content cheaper to produce in terms of time and labor. Google has also been clear that the method of production isn’t the point; the point is whether the content is helpful and created for people. The issue is what “helpful” looks like when the internet is flooded with fluent, same-shaped pages.

When output becomes easy, the bottleneck moves to trust and differentiation.

That sounds like branding fluff until you translate it into SEO mechanics. Differentiation shows up as information gain, unique experience, original data, clear point of view, and editorial standards that prevent you from publishing a page that could have been generated by anyone else. Trust shows up in consistency, accuracy, and a site that doesn’t feel like it’s trying to win by volume.

This is also where “it depends” gets sharper. AI is great for drafting, outlining, summarizing internal docs, repurposing webinars, and improving consistency across large content sets, if your inputs are strong and your strategy is clear. It can help a mature team move faster without lowering standards.

AI is dangerous when it papers over weak strategy, unclear intent, or missing expertise. In that scenario, it doesn’t just produce mediocre pages. It produces a lot of them. And they start competing with each other. You get same-but-different pages that dilute topical focus, confuse internal linking, and make your site’s “aboutness” harder for search systems to resolve.

Meanwhile, Google has simultaneously tightened language around scaled content abuse: the problem isn’t automation, it’s automation used to generate lots of unoriginal, low-value pages for ranking manipulation. That policy framing matters because it maps directly onto what sloppy AI publishing looks like from the outside: large amounts of content that feels mass-produced, even if nobody intended it to be “spam.”

This is the uncomfortable middle: you can use AI and be fine. You can also use AI and accidentally build a machine for manufacturing pages your own team wouldn’t choose to read.

AI reporting tools are getting good at turning dashboards into sentences. That’s useful. It’s also where teams can quietly lose discipline.

The core issue is that AI is good at summaries and bad at causality. It will confidently tell a clean story using the most available explanation, not the most correct one. If rankings dropped after a deployment, it will blame the deployment. If traffic rose after you published new pages, it will praise the new pages. It is, by design, a pattern completer—not a hypothesis tester.

This gets worse when the underlying data has blind spots or delays.

Even Google’s own tooling can lag at times. For example, Google confirmed that Search Console’s page indexing / index coverage reporting can experience delays that affect what you see in the UI, while crawling and indexing continue as normal. If your AI-generated “insights” don’t account for that kind of reporting artifact, you get what I think of as correlation laundering: a narrative that sounds evidence-based because it references metrics, but is still just storytelling.

The fix isn’t to avoid AI in reporting. The fix is to use AI where it’s strong and keep humans responsible for what matters. Let AI synthesize across Search Console, analytics, logs, and release notes. But require a human-owned hypothesis and a measurement plan before you treat the story as true.

“It depends” isn’t a dodge. It’s a clue that you haven’t identified the binding constraint.

So replace “it depends” with: it depends on which constraint is binding right now.

Is crawl capacity the limiter, because your site is huge and your URL ecosystem is noisy? Is index quality the limiter, because duplication and thin variants are crowding out the pages you actually care about? Is content differentiation the limiter, because you’re publishing answers that already exist everywhere? Is site trust the limiter, because quality is inconsistent and the brand doesn’t feel reliable? Or is internal incentive the limiter, because speed is rewarded more than outcomes?

A quick way to triage this without hand-waving is to force four questions in the open. What are we trying to improve: visibility, conversions, coverage, or efficiency? What could break at scale: crawl budget, duplication, brand trust, or governance? What signals are we actually changing: internal linking, entity clarity, page quality, engagement proxies? What do we need to measure to know we’re right, and how long will it take before the measurement is meaningful?

This framework isn’t glamorous. That’s why it works. It pulls the conversation back to systems.

Treat AI as a multiplier, not a shortcut.

If you’re an in-house team trying to use AI without creating a future cleanup project, the practical moves are boring on purpose. Put guardrails on publishing so scale doesn’t turn into index bloat, starting with templates, dedupe rules, indexation policies, and QA checks tied to real failure modes. Raise the content bar so every page has a reason to exist that you can say out loud: new information, proof, experience, or a point of view that isn’t interchangeable. Upgrade reporting discipline so AI can write the summary, but humans own the hypothesis, the pre/post measurement, and the decision.

The punchline is simple. AI won’t replace SEO. It will expose whether your SEO is infrastructure or vibes. But, don’t worry, this newsletter is here to help you figure out the difference.

Natural Language Processing (NLP): The field of computer science focused on understanding and generating human language. Learn More

One more thing: AI is only as good as it’s operator, and if you are reading this newsletter, you are better than most!

Till next time,

PS: Let me know what you think of this issue, or anything else here: [email protected]