How AI Search Is Changing SEO (And What Still Works)

What's genuinely changing

The mechanics of search haven't changed as much as the interface has. People still have questions and still want answers. What's different is how many of them are willing to click through a page of results to get there, versus just reading what an AI tool hands them directly.

A meaningful chunk of query volume that used to land on Google is now starting — or ending — inside ChatGPT, Perplexity, Claude, or Google's own AI Overviews. Someone asks "what's the best invoicing tool for freelancers" and gets three names and a short rationale, no tab-switching required. For a certain category of query — comparisons, definitions, quick recommendations, "how do I" questions — the round trip through ten blue links is starting to look unnecessary to the person asking, and they're voting with their behavior.

The practical effect is fewer clicks per query, even when your content is the thing being summarized. This is the zero-click shift people talk about: your page can be the source of the answer and still receive no visit, because the answer already satisfied the person. That's a real change in how traffic gets attributed, not a hypothetical one.

The other shift is in how visibility gets earned. Classic SEO ranking was largely a contest over a query and a URL — you optimized a page to rank for a phrase. AI answer engines don't work off a single ranked list; they synthesize from multiple sources and decide, query by query, who gets cited or mentioned and who doesn't. Being the best-optimized page for a keyword matters less than being a source the model trusts enough to pull from or reference by name. That's a different game, closer to earning a reputation than winning a slot.

It's also worth naming what this isn't. It isn't the end of organic discovery, and it isn't a sign that content no longer matters. People still research, compare, and read before they buy something, especially anything with a real price tag or a real switching cost attached. What's changing is the shape of the first few steps in that research — more of it happens inside a conversation with an AI tool before a person ever lands on a website, and by the time they do land somewhere, they've often already formed an opinion based on what that tool told them. That puts a premium on being part of the answer, not just a destination at the end of it.

What hasn't changed

It's tempting to treat this as a total reset, but most of the underlying mechanics are intact. AI answer engines still depend on the same raw material search engines always depended on: content that can be crawled, understood, and judged as credible. None of that went away — if anything, it matters more, because now two different kinds of systems need to parse your site instead of one.

A few things that are just as true now as they were before any of this started:

• Crawlability still matters. If a bot can't fetch and parse your page — broken robots.txt rules, content locked behind JavaScript that doesn't render for crawlers, pages with no clean HTML structure — it can't be cited, summarized, or ranked by anything, AI or otherwise.
• Credible, specific content still matters. Answer engines favor sources that say something clear and verifiable over sources that pad a page with vague claims. Thin, generic content was already a losing bet under old-school SEO. It's a losing bet here too, just for a slightly different reason.
• Backlinks and mentions still matter. Being referenced by other credible sites is still one of the strongest signals of trustworthiness a page can have, whether the thing reading that signal is a search ranking algorithm or a language model deciding what to cite.
• Technical fundamentals still matter. Page speed, clean markup, working sitemaps, accurate metadata, mobile rendering — the boring infrastructure work is still table stakes. None of the new tools reward a broken foundation.

The honest way to describe this moment is not "SEO is dead," it's "the scoring changed, but the inputs it scores didn't." If your content strategy up to now has actually been about being useful and well-built, you're closer to fine than the more panicked takes suggest.

Why this matters more when you're small

A large company can afford to chase every channel at once and let the data sort out what worked. A two-person startup can't. Every hour spent writing content is an hour not spent on product, sales calls, or the dozen other things founders are supposed to be doing. That makes the question of where content effort goes a real strategic decision, not a nice-to-have.

The risk with this transition isn't that AI search exists — it's spending your limited content budget chasing tactics that only made sense under the old ranking game. Keyword-stuffing a page for a search term nobody types into ChatGPT, publishing thin listicles purely to occupy rankings, or optimizing exclusively for click-through rate on a results page that's shrinking — those are the tactics most exposed by this shift, and they were never durable to begin with.

What's durable is being a source worth citing: clear explanations, real opinions backed by reasoning, content that answers the actual question instead of circling it for SEO padding. That kind of content holds up whether the reader is a human scrolling results or a model deciding what to reference. For a startup with limited hours, that's the argument for spending them on substance over volume — fewer pages, each one actually worth finding, rather than a pile of pages built to rank for search terms that increasingly get answered without a click at all.

There's a second, quieter reason this matters for a small company specifically: distribution outside your own website is no longer optional groundwork, it's a direct input into whether AI tools know you exist at all. A large, well-funded competitor can win visibility partly through sheer volume of content and links accumulated over years. A startup usually can't out-publish anyone. What a startup can do is be genuinely present and useful in the smaller number of places its actual customers already gather — a specific subreddit, a specific Slack community, a specific corner of X or LinkedIn — and let that specificity be the advantage. Answer engines pick up on real engagement and real mentions in those spaces even when the total volume is modest, because relevance and credibility count for more than raw quantity.

What to actually do differently, starting now

None of this requires abandoning what you were doing before. It requires being more deliberate about which parts of it you keep doing and adding a few habits that specifically help with how AI tools find and use your content.

1. Write content that answers the question in the first two sentences. Don't bury the direct answer under three paragraphs of throat-clearing. Answer engines pull from content that states its point plainly; so do impatient humans.
2. Make your pages easy to parse, not just easy to read. Use real headings, short paragraphs, and structured lists where they fit. Confirm your important pages render without JavaScript, or at least that a crawler can get the content some other way.
3. Get mentioned in places that already have credibility. A mention or link from a site or community that's trusted in your space does double duty — it can drive a referral click and it's the kind of signal AI systems weigh when deciding what to cite. This is where consistent participation on places like Reddit, LinkedIn, Hacker News, or Quora pays off over time, not because you're gaming a channel but because genuine, useful answers in those places tend to get referenced elsewhere.
4. Say something instead of summarizing everyone else. If your article restates what the top five existing results already say, there's no reason for a model — or a person — to prefer it. Bring a specific opinion, a specific number from your own experience, or a specific example.
5. Check whether AI tools already mention you, and how. Ask ChatGPT or Perplexity questions your prospective customers would ask and see if your product shows up, what it's described as, and what it's compared to. That tells you more about your actual visibility right now than a keyword rank ever will.
6. Don't drop the fundamentals to chase the new thing. Keep your sitemap current, keep pages fast, keep fixing broken links. This work is unglamorous and it's still the floor everything else stands on.

Most of this is just consistent execution — publishing content that's actually worth citing, showing up in the channels where your audience already has conversations, and tracking whether that effort shows up in AI-generated answers over time. That's the part that's easy to know and hard to keep doing every week alongside everything else running a startup. This is one of the areas Wally is built for: it researches where your audience already talks about problems you solve, drafts posts and replies for channels like Reddit, LinkedIn, X, Hacker News, and Quora, and helps get consistent, credible content in front of both people and the tools now answering their questions — all queued for your review, nothing published without approval.