The question gets asked in a way that cannot be answered. "Do AI-built websites rank?" treats the builder as the variable, when the builder is not what Google evaluates. Google evaluates whether a page can be crawled, whether it is the best answer to a query, and whether anything on it is worth returning. An AI builder produces some of what that requires, produces a rough version of some more, and produces none of the rest. Knowing which bucket each requirement falls into is the entire answer.
Yes, a site built with an AI builder can rank, and Google does not penalise a page for having been generated by AI. Google's own guidance states that its focus is the quality of content rather than how it was produced, and its spam policies target scaled content abuse regardless of method. What holds AI-built sites back is narrower and more fixable: client-side rendering that hides content from crawlers, headings and structured data written for appearance rather than semantics, no information architecture mapped to real search demand, no redirect map when replacing an old site, and content that contains nothing a search engine could not already find elsewhere.
What Google actually says about AI-generated sites
This needs settling before anything else, because most of the anxiety about AI-built sites is aimed at the wrong target. Google's Search Central documentation on generative AI content states that appropriate use of AI is not against its guidelines and that its ranking systems reward original, helpful content, however it is produced. The relevant restriction lives in the spam policies, under scaled content abuse: generating many pages primarily to manipulate rankings rather than to help people. That policy is deliberately method-agnostic. AI, human or scraped, it applies the same way.
So the honest framing is not "will Google catch that I used AI". It is "did the tool produce a site that deserves to rank". Those are completely different questions, and the second one is answerable page by page.
The ten requirements, and which ones an AI builder actually delivers
Below is every component that materially decides whether a small business site earns organic traffic, matched against what AI builders typically produce. The pattern is consistent: the technical scaffolding is mostly handled, the strategic layer is almost entirely absent, and the middle band is where the expensive surprises live.
| Requirement | Typical AI-builder output | What it costs you if it is missing |
|---|---|---|
| Content present in the server-rendered HTML | Varies sharply by tool. App-style builders often render body copy client-side | Googlebot renders JavaScript, but in a second pass that can delay indexing. AI answer-engine crawlers are generally weaker at rendering than Googlebot, so client-side content can be effectively invisible to them |
| Indexable URLs, clean canonicals | Usually handled, but duplicate or missing canonical tags are a common audit finding | Split signals across near-identical URLs, and pages competing with themselves |
| Title tags and meta descriptions written per page intent | Generated, frequently duplicated across pages | Weak click-through, and Google rewriting your titles because yours describe nothing specific |
| Semantic heading hierarchy | Headings written for how they look and read rather than what they are. This is the single most consistent finding in published audits of AI-built sites | Passage-level understanding degrades, which hurts both ranking and extraction into AI answers |
| Structured data matching visible content | Inconsistent. Some builders inject schema after a delayed client-side render, so some crawlers see it and others do not | No rich results, and validation tools that disagree with each other, which makes the problem hard to even detect |
| Information architecture mapped to real search demand | Not produced. The tool builds a plausible site structure, not one derived from what people search | The largest single loss. Pages exist for topics nobody searches, and no page exists for the queries that would convert |
| One page per search intent, no cannibalisation | Not produced, and AI page generation makes overlap more likely rather than less | Two or three of your own pages splitting the same query, none of them winning it |
| Internal linking that expresses topical relationships | Navigation links only. Broken internal links are a routine finding | Orphan pages, and no signal to Google about which of your pages is the authority on a topic |
| Information gain: something the ranking pages do not already say | Not produced. Generic, broadly correct, widely available copy is the default output | No reason for Google to rank you above the incumbents, and nothing for an AI answer engine to cite |
| Redirect map when replacing an existing site | Never produced | The most expensive failure on this list. Existing rankings and links point at URLs that now return 404 |
Read the bottom half of that table as the actual answer to the headline question. Nothing in it is a fault of AI specifically. These are the parts of a website build that require knowing your market, your existing traffic and your competitors, and a tool that has seen none of those things cannot produce them. That is also why "I will just prompt it better" does not close the gap: the missing inputs are data about your business, not instructions.
If you are replacing an existing site, deal with the redirect map before anything else on this page. Export every URL that has received organic traffic or an inbound link in the last twelve months, map each to its closest equivalent on the new site, and test the redirects before launch, not after. This is the one item on the list where the damage is retroactive and largely irreversible.
A four-step diagnostic you can run on your own site today
You do not need an audit tool to find most of this, and audit tools are unreliable here for a specific reason: many read the raw HTML while Google reads the rendered page, so a tool can report a missing H1 that is present, or report schema that only some crawlers ever see. Check the rendered and the raw version separately.
- Compare raw HTML against the rendered page. Pick a distinctive full sentence from a body paragraph. View source, then search for it. If the sentence is not in the source, your content is arriving via JavaScript. Google will usually still get there; AI crawlers often will not.
- Use URL Inspection in Search Console, then open the HTML tab. This is what Google actually fetched and rendered. Check your main heading, your body copy and your structured data are all present in it. If the tool and your browser disagree, trust the tool.
- Compare pages published against pages indexed. The Pages report in Search Console gives you indexed and not-indexed counts with reasons. A large not-indexed bucket labelled "Crawled – currently not indexed" on a small site is usually a quality and duplication signal rather than a technical one.
- Read the top three results for the query you most want. Then ask what your page contains that those three do not. If the answer is nothing, the problem is not technical and no amount of technical work will fix it. This is the test most site owners skip, and it is the one that predicts outcomes best.
Step four is worth dwelling on, because it is where AI-built sites most often fail on their own terms. A tool that writes your service page from a prompt is drawing on the same widely-available material as every competitor's page. The output is usually accurate and usually unremarkable, and unremarkable is a ranking problem rather than a quality problem. Our technical SEO audit checklist covers steps one to three in more depth; step four is a content decision, not an audit item.
Why this matters more for AI search than for Google
There is a second-order effect that makes the information-gain problem sharper than it used to be. Answer engines select sources to cite, and a page that restates the consensus gives them no reason to cite it over the page they already have. Generic output is not just hard to rank, it is hard to quote.
The gap also runs the other way, and it is worth knowing before you set expectations: being cited by an AI system is not the same as being recommended by it. An analysis of B2B software queries in Google AI Overviews found that when a brand's own page was cited, the AI Overview recommended a different brand in roughly 69% of those cases. We worked through what closes that gap in cited but not recommended. The relevance here is that both citation and recommendation depend on the page saying something specific and attributable, which is exactly what a generated first draft does not do.
Two practical consequences. First, if your content arrives client-side, fix that before you worry about AI visibility at all, because a crawler that cannot render your page cannot consider it. Second, treat generated copy as a structural draft and add the parts only you have: your pricing, your process, your constraints, the trade-offs you actually talk customers through. The AEO and GEO guide covers the formatting side of the same problem.
Where an AI builder is genuinely the right choice
The honest version of this argument has to include the cases where the tool wins, and there are several. An AI builder is a good decision when the site's job is to exist, be found by people who already know your name, and answer a short list of questions. Specifically:
- You have no site at all and no budget. A competent AI-built site beats no site and beats a stalled project, and it can be replaced later. Speed has real value when the alternative is nothing.
- Your demand is local and driven by your Google Business Profile. If most enquiries arrive through maps and calls, the site is a credibility check rather than an acquisition channel, and the strategic layer in the table above matters much less.
- You are validating an offer. Spending twelve weeks and five figures on a site for a proposition you have not tested is the more expensive mistake.
- The site is under about ten pages and will stay that way. Cannibalisation, internal linking and architecture are problems of scale. They barely apply at that size.
The pattern across all four is that organic search is not the acquisition channel. Once it is, the missing items in the table stop being theoretical, because each one is a reason a competitor outranks you. If you are weighing platforms rather than the AI question specifically, the builder comparison is the more direct piece, and how AI is changing what a web development company delivers covers the same technology from the agency side.
Fix it or rebuild it? A decision rule
If you already have an AI-built site that is not earning traffic, the question is whether to remediate or start again. The answer turns on two things only.
| Is the URL structure and page inventory roughly right? | Is the content specific to your business? | Do this |
|---|---|---|
| Yes | Yes | Remediate. Fix rendering, headings, schema, internal links and titles. You are keeping something worth keeping |
| Yes | No | Keep the structure, rewrite the content page by page in priority order. The architecture is the expensive part and you have it |
| No | Yes | Restructure with redirects. Map the new architecture from search demand first, then move the existing content into it |
| No | No | Rebuild, and this time start from a keyword and intent map rather than from a prompt. Remediation here means paying to preserve nothing |
Most sites land in the second row, which is the encouraging one: the structure survives, and the work is a prioritised rewrite rather than a rebuild. Prioritise by commercial value, not by page order. Your two or three highest-intent service pages first, then the pages that support them, and leave the rest until those are done. Our SEO and web design playbook sets out how that sequence should run alongside design work rather than after it.
The measurement that tells you whether it worked
Set the baseline before you change anything, because the most common outcome of remediation work is a genuine improvement that nobody can prove. Record four numbers today: impressions and clicks from Search Console for the last 28 days, the count of indexed versus not-indexed pages, the number of distinct queries returning any impression, and your organic conversions. Then re-check at six and twelve weeks.
Query count is the most useful of the four and the least watched. A site with an architecture problem typically shows impressions concentrated on a handful of queries, mostly branded. When the architecture starts working, the distinct query count grows before clicks do, often by weeks, so it is the earliest honest signal that the work is landing. Clicks follow, and revenue follows that.
Frequently asked questions
Does Google penalise websites built with AI?
No. Google's Search Central guidance on generative AI content states that appropriate use of AI is not against its guidelines, and that its ranking systems focus on the quality and originality of content rather than on how it was produced. The relevant restriction is the scaled content abuse policy in Google's spam policies, which targets generating many pages primarily to manipulate rankings rather than to help people. That policy applies the same way to AI-generated, human-written and scraped content.
Why does my AI-built website get no traffic?
In most cases for one of four reasons. The site's structure was generated from a plausible template rather than mapped to what people actually search, so no page targets a query worth winning. The content is broadly correct but says nothing the already-ranking pages do not, giving Google no reason to prefer it. The body copy arrives via JavaScript, which delays Google's indexing and can make the page invisible to AI answer-engine crawlers. Or the site replaced an older one without a redirect map, so previous rankings and links now point at 404s.
Are AI website builders good for SEO?
They handle the technical scaffolding reasonably well and the strategic layer barely at all. Indexable URLs, sitemaps, mobile layout and basic metadata are usually produced. Semantic heading hierarchy, structured data that matches visible content, and non-duplicated titles are produced inconsistently, with headings being the most common consistent weakness in published audits. Keyword and intent mapping, internal linking that expresses topical relationships, redirect planning and original information gain are not produced at all, because they depend on data about your market that the tool does not have.
How do I check whether my site's content is visible to search engines?
Pick a distinctive full sentence from a body paragraph, view the page source and search for it. If it is absent, the content is rendered client-side. Then open URL Inspection in Google Search Console and view the HTML tab, which shows what Google actually fetched and rendered, and confirm your heading, body copy and structured data all appear there. Where an audit tool and your browser disagree, the audit tool is probably reading raw HTML while Google reads the rendered page.
Should I fix an AI-built site or rebuild it?
It depends on two things: whether the URL structure and page inventory are roughly right, and whether the content is specific to your business. If both are true, remediate. If the structure is right but the content is generic, keep the structure and rewrite page by page in order of commercial value, which is the most common situation. If the structure is wrong but the content is good, restructure with redirects. If both are wrong, rebuild starting from a keyword and intent map, because remediation would mean paying to preserve nothing.
When is an AI website builder the right choice?
When organic search is not your acquisition channel. Four cases fit clearly: you have no site and no budget, and something beats nothing; your demand is local and arrives through your Google Business Profile, so the site is a credibility check; you are validating an offer that has not been tested; or the site is under about ten pages and will stay that way, which is below the scale at which architecture and cannibalisation matter. Once organic search becomes a real channel, each missing requirement turns into a reason a competitor outranks you.
The takeaway
Google does not rank or penalise a page for how it was made, so stop asking whether AI-built sites rank and start asking which of the ten requirements yours actually meets. The technical scaffolding is usually present, the semantic layer is usually rough, and the strategic layer, architecture mapped to real search demand, one page per intent, internal linking and genuine information gain, is usually absent because the tool never had the inputs to produce it. Run the four-step diagnostic, fix rendering first and redirects before launch, then rewrite your highest-intent pages so they say something a competitor's page does not. Watch distinct query count as the early signal that it is working.
Sources & further reading
- Google Search Central, “Google Search's guidance about AI-generated content”
- Google Search Central, “Spam policies for Google web search” (scaled content abuse)
- Nikki Pilkington, “Are AI-built websites any good for SEO? I audited 5 to find out”
- Google Search Central, “Understand JavaScript SEO basics”