For years, “duplicate content” has been SEO shorthand for a technical housekeeping issue. Fix your canonicals, consolidate your URLs, tidy up your syndication, dot your ‘i’s, cross your ‘t’s, and move on. By now, most experienced SEOs have that down to a fine art.

The problem is, that old-school interpretation of duplicate content is no longer the deciding factor in whether your page gets seen.

AI search is changing how people find information, but it’s also defining how many different versions of that information the internet is allowed to have.

In traditional search, a single results page might serve up ten similar links on the same topic. AI search doesn’t give ten answers. It gives one. And if your content isn’t cited within that one answer, it’s invisible, even if your site is technically well-optimised.

At SALT, we call this “the new duplicate content problem”: sameness across the web, not just duplicate URLs on your own site.

Ranking for AI search has turned duplicate content from a technical box-ticking exercise into a strategic conundrum. Here’s what you should do about it.

The old problem: Duplication within your own site

The traditional understanding of duplicate content is sort of like a plumbing issue. Somewhere behind the walls of your site, the same content is flowing out through multiple pipes instead of one, and search engines can’t work out which is which.

Common causes include:

  • URL variations: the same page is accessible via multiple parameters, tracking tags, or trailing slashes, and each one is treated as a separate page.
  • Printer-friendly or app versions: duplicate copies of a page have been created for a different format or device.
  • Ecommerce variants: there are multiple near-identical product pages differentiated only by colour, size, or SKU, with little unique content between them.
  • Syndicated content: the same article has been republished across multiple sites or sections without a clear canonical source.

In this interpretation of duplicate content, the damage is real but contained. The duplication dilutes ranking signals across multiple URLs instead of consolidating them behind one strong page. In plumbing terms, your water pressure is too low because one supply is being split too many ways (bear with us, this analogy does go somewhere).

This wastes crawl budget, so search engines spend less time discovering your genuinely new or updated content. And in the worst cases, it creates the impression of a thin or low-effort site, denting trust in the domain as a whole.

The fix, however, is equally contained: canonical tags, redirects, better site architecture, clearer content governance. Solve the technical problem, and improved visibility tends to follow. That’s why so far, duplicate content has largely been the remit of technical SEOs and developers, rather than a boardroom issue.

In the AI search era, that cosy containment no longer exists.

The new problem: Duplication across the entire internet

You can’t plumb your way out of this one. In AI search, there might still be a leak in your walls, but it’s also happening across the whole street.

To clarify, this isn’t about AI tools being used to mass-produce cookie-cutter content, though that doesn’t help. It’s about how AI search cites (or more accurately, doesn’t cite) content that already sounds like everyone else’s, regardless of who or what wrote it.

Instead of returning a list of blue links for a query, AI search engines provide one synthesised answer, built from whichever sources the model deems most useful and trustworthy. Traditional rankings aren’t a reliable guide to which sources get chosen.

The sources Google AI Mode cites often differ from the classic top 10 for the same query, so treat organic results as a weak proxy for AI visibility. Start with a structured competitor analysis in AI search to see which sources are cited for your priority queries, and where you’re missing.

If three, ten, or a hundred sites are saying broadly the same thing about a topic, AI doesn’t need to cite all of them. It just needs to cite those saying it best, or failing that, saying it first.

Being understood still matters. Clear entity signals help AI systems understand who you are and what you do. Being understood isn’t the same as being chosen when a hundred sites say the same thing. At that point, sameness is the bottleneck, not a missing entity signal.

That means if you want to perform in AI search, the definition of “duplicate content” has expanded from your own site to every other site on the internet covering the same topic.

If your guide to a topic reads like a well-organised summary of what’s already publicly available, an AI search model can produce that summary too, without ever needing to cite your page.

You’re not just competing with other companies for a ranking position anymore. You’re competing with an AI model’s ability to generate content you’ve already published.

Originality: The new visibility signal

This is why originality is so important to AI search visibility. That doesn’t necessarily have to mean “written completely from scratch,” but it does have to mean “contains something that didn’t already exist anywhere else.” That’s the difference between content AI still needs you for, and content AI can just produce itself.

Originality can come from:

  • Original research: studies or surveys nobody else has access to.
  • First-party performance data: real results and outcomes from your own work.
  • Genuine subject-matter expertise: a unique, informed point of view, not a rehash of what’s already been said.
  • Direct customer and market insight: what your business sees that competitors don’t.

An AI model can synthesise, summarise, and rephrase what’s already out there, but it can’t generate information that doesn’t exist yet.

Original content built on real, first-party insight is catnip for AI because, until you hit publish, there’s nothing else out there quite like it. So, when someone asks a question that touches on your niche, AI has nobody else to cite but you.

On the other hand, content built solely on existing public consensus is vulnerable, because that’s exactly what AI already excels at distilling into a single answer.

The same goes for templated copy generated at scale. If it’s built from what’s already public, it adds nothing an AI model can’t synthesise and produce itself. First-party data provides the difference.

The research points the same way. Where citation patterns have been measured, original research and first-party data tend to earn AI citations more often than interchangeable guides. Focus your content calendar on content built around new data, information, and insights rather than summary posts anyone could rewrite.

For the C-suite, this reframes the risk. Underperforming organic content isn’t necessarily a keyword or technical problem; it may be that the content itself has become interchangeable. This is the argument that gets budget redirected from more content to better content, and it’s backed by how AI systems actually select sources.

Next steps: Run a duplicate content audit, the new way

Previously, duplicate content audits aimed to answer the question: “Are all our pages indexed correctly, and do they compete with our other pages?”

That audit still has its place, but it’s no longer sufficient on its own.

Duplicate content audits should now ask a sharper question: “Would an AI model have already known everything on this page before we published it?” Or, put another way: “Could AI produce this article tomorrow, using only what’s already publicly available?”

If the honest answer is yes, that page is vulnerable, regardless of how well it’s written, how clean its technical SEO is, or how well it once ranked in traditional search.

Here’s a quick step-by-step guide to running a duplicate content audit, the new way, with AI search as a priority:

  1. Inventory your highest-traffic and highest-intent pages: Prioritise the content most exposed to AI-driven answer generation, such as informational and comparison content.
  2. Score each page for originality: Ask whether each page contains data, insights, or perspectives that exist nowhere else. Run a named-system check by asking the page’s core question in ChatGPT, Gemini, and Google AI Mode on the same day. Then compare each answer with your page. If they already cover your points, treat the URL as exposed.
  3. Identify your proprietary assets: Check if you have customer data, case studies, or internal expertise sitting underused.
  4. Rebuild exposed content around what only you know: This might mean fewer but more in-depth pieces of content, organised as content hubs: topic clusters that reinforce each other with long-form, well-structured depth on subjects you can genuinely own, rather than a wide range of content covering the same ground as everyone else.
  5. Treat this as an ongoing discipline, not a one-off project: as AI search evolves, so will the criteria for what counts as original enough to be cited.

Duplicate content hasn’t disappeared. It has evolved

OK, so we might not have been completely candid in this article.

The truth is, duplicate content was never really about URLs and canonical tags (or about pipes and pressure). It has always been about giving search engines a reason to choose your page over others. Those reasons have just changed.

Since AI search exploded in popularity, your content’s never-ending battle against every other page like it has simply been made explicit, immediate, and far larger in scale.

As AI-driven discovery continues to take a larger share of how people find answers, the businesses that still treat duplicate content as a canonicals ticket will find themselves losing visibility to those that treat sameness as a strategic problem.

There’s a new duplicate content problem – but it’s solvable, starting with the question: “Would AI have known this before we published it?”

Let SALT.agency help you create content AI can’t ignore

If you need an AI search strategy built around genuine expertise, SALT.agency can help you create unmissable content and give AI a reason to cite your brand.

Get in touch with our team to arrange a discovery session and find out how we can improve your visibility in AI search.