Key takeaways
- A Semrush study found only ~35% URL overlap between Google AI Mode citations and traditional search results, meaning most cited pages don't rank highly in organic search.
- Google AI Mode uses a "query fan-out" process, breaking one search into multiple sub-queries and pulling from different sources for each, which traditional ranking signals don't account for.
- AI Mode rewards content that directly answers questions, demonstrates topical depth, and is structured for machine parsing -- not just content that has accumulated backlinks.
- Monitoring your AI Mode citations separately from your traditional rankings is now a practical necessity, not optional.
If you've been assuming your top-ranking pages are the ones Google AI Mode cites, you're probably wrong. Not slightly off -- significantly wrong.
A Semrush study comparing AI Mode citations to traditional search results found roughly 54% domain overlap but only about 35% URL overlap. That gap between domain and URL tells an interesting story: Google might trust your site in general, but it's often choosing different pages on that site than the ones you've worked hardest to rank. And for many brands, it's choosing competitor pages entirely.
This isn't a bug or a temporary quirk. It reflects something structural about how AI Mode works -- and understanding it is the difference between having a search strategy that holds up in 2026 and one that's quietly losing ground.
What Google AI Mode actually does differently
Traditional Google search is, at its core, a retrieval and ranking system. You type a query, Google matches it against its index, applies hundreds of ranking signals, and returns a list. The pages at the top are the ones that best satisfy that specific query according to those signals.
AI Mode does something more complex. When you submit a query, it doesn't just look for the best single answer. It runs what Google calls a "query fan-out" -- breaking your original question into multiple sub-queries, running those in parallel, pulling results from different sources for each, and then synthesizing everything into a coherent response.
So if someone asks "what's the best project management software for remote teams under 10 people," AI Mode might simultaneously run sub-queries about pricing, team size suitability, remote collaboration features, and user reviews. Each of those sub-queries might pull from a different page, a different domain, or even a different type of content (a comparison article, a Reddit thread, a product page).
Your #1 ranking for "best project management software" doesn't automatically win any of those sub-queries. What wins each one is the page that best answers that specific sub-question.
Why traditional ranking signals don't translate cleanly
Traditional SEO ranking signals -- domain authority, backlink profiles, keyword density, click-through rates -- were built to predict relevance for a specific query. They work reasonably well for that purpose.
But AI Mode isn't just asking "which page is most relevant to this query?" It's asking something closer to "which page contains the most useful, trustworthy, and directly applicable information for this specific aspect of what the user wants to know?"
That's a subtle but important difference. A few things fall out of it:
Backlinks matter less at the page level. A page with 500 referring domains might rank #1 organically but get ignored by AI Mode if it doesn't directly answer the sub-question being evaluated. Meanwhile, a page with 12 referring domains but a clear, specific answer to a niche question might get cited repeatedly.
Freshness matters more than it used to. AI Mode appears to weight recency more heavily for certain query types, particularly anything involving current events, product comparisons, or evolving topics. A well-optimized page from 2022 can lose out to a less-optimized page from last month.
Structural clarity is now a ranking factor. Pages that use clear headers, answer questions directly in the first paragraph, include structured data, and avoid burying the key information in long preambles tend to get cited more. AI systems are parsing your content for extractable answers, not reading it the way a human would.
Topical depth beats keyword targeting. A page that covers one topic thoroughly -- including related questions, edge cases, and comparisons -- is more likely to satisfy multiple sub-queries than a page optimized for a single keyword. This is why some sites with relatively modest traditional SEO metrics punch above their weight in AI Mode.
The query fan-out problem, explained with an example
Say you run a financial planning firm and you've spent years building a page that ranks #2 for "retirement planning for self-employed." Solid backlinks, good on-page SEO, decent click-through rate.
When someone asks Google AI Mode "how should a freelancer plan for retirement," the fan-out might generate sub-queries like:
- What retirement accounts are available to self-employed people?
- How much should a freelancer contribute to retirement?
- What are the tax implications of SEP-IRA vs Solo 401k?
- What are common mistakes freelancers make with retirement planning?
Your page might answer one or two of those well. But if a competitor has a dedicated page on SEP-IRA vs Solo 401k comparisons, and another site has a detailed piece on freelancer tax strategy, AI Mode might cite both of them instead of your more general page -- even though your page ranks higher for the broad query.
This is why people are seeing AI engines cite pages that rank nowhere on Google. Those pages aren't ranking for the broad query, but they're the best answer to a specific sub-question that AI Mode cares about.
The content structure gap
One thing that consistently surprises people when they audit their AI Mode citations: the pages that get cited often look different from typical top-ranking pages.
Cited pages tend to:
- Answer a specific question in the first 100 words
- Use headers that mirror how people phrase questions (not just keyword-stuffed headers)
- Include concrete numbers, comparisons, or step-by-step instructions
- Avoid heavy reliance on introductory fluff before getting to the point
- Have clear entity relationships -- they're explicitly about a specific thing, not vaguely related to a topic
Pages that rank well traditionally but get skipped by AI Mode often:
- Lead with brand storytelling or general context before getting to the answer
- Target a broad keyword rather than a specific question
- Rely on authority signals (backlinks, brand recognition) rather than content specificity
- Are optimized for human reading patterns rather than machine extraction
This doesn't mean you need to rewrite everything. But it does mean that the content decisions you make going forward should account for how AI systems parse and extract information, not just how humans read it.
Google AI Mode vs. AI Overviews: they're not the same thing
It's worth clarifying something that causes a lot of confusion: Google AI Mode and Google AI Overviews are different products with different citation behaviors.
| Feature | AI Overviews | AI Mode |
|---|---|---|
| Appears in | Standard search results (above organic) | Dedicated AI Mode tab |
| Query coverage | Selective (informational queries) | Broad (most query types) |
| Citation overlap with organic | Higher (~60-70% URL overlap) | Lower (~35% URL overlap) |
| Depth of response | Summary-level | Multi-turn, conversational |
| Sub-query behavior | Limited fan-out | Aggressive query fan-out |
| Freshness weighting | Moderate | Higher |
AI Overviews stay closer to traditional rankings because they're designed to supplement a standard search results page. AI Mode is more autonomous -- it's meant to replace the need to click through to multiple pages, so it casts a wider net and makes more independent retrieval decisions.
If you're tracking your visibility in AI Overviews and assuming that tells you how you're doing in AI Mode, you're missing a significant part of the picture.
What this means for your content strategy
The practical implication is that you need to think about content in two dimensions now: traditional ranking signals and AI citation signals. They overlap, but they're not the same.
A few things worth doing:
Audit what's actually getting cited. Before changing anything, find out which of your pages are appearing in AI Mode responses and which aren't. You might find that a mid-tier blog post is getting cited more than your cornerstone content. That's useful information.
Map your content against sub-questions, not just head terms. For each major topic you cover, think about the specific questions someone might ask within that topic. If you don't have a page that directly answers each of those questions, you have a gap.
Improve answer clarity on existing pages. You don't always need new content. Sometimes adding a clear, direct answer in the first paragraph -- before any context-setting -- is enough to make a page more citable.
Use structured data. FAQ schema, HowTo schema, and Article schema all help AI systems understand what your content is about and extract specific answers. This is one of the clearest signals you can send.
Watch your competitors' citations, not just their rankings. If a competitor is getting cited in AI Mode for queries you care about, look at what page is being cited and why. It's often more instructive than looking at their domain authority.
Tools like Promptwatch can help here -- it tracks which pages are being cited by AI engines, shows you the specific prompts driving those citations, and identifies gaps where competitors are visible but you're not.
For broader SEO and AI visibility tracking, Semrush has published useful research on AI Mode behavior and includes some AI visibility features in its toolkit.
If you want dedicated AI search monitoring without the full SEO platform overhead, Otterly.AI and Peec AI are worth looking at for tracking citation patterns.
The false assumption that's costing brands visibility
There's a working assumption baked into most marketing teams' thinking: if we rank well on Google, we're visible. That assumption made sense for a long time. It's increasingly wrong.
Google AI Mode is now available in over 200 countries. It's not a beta feature or a niche use case -- it's becoming the default search experience for a growing share of queries. And its citation logic is operating with significant independence from the traditional ranking algorithm.
As one LinkedIn post from a search strategist put it: "Google's AI citation logic is increasingly disconnected from its own ranking algorithm. Which means brands are operating on a false assumption."
That's not hyperbole. The Semrush data backs it up. A 35% URL overlap means that roughly two-thirds of the pages AI Mode cites are not the pages you'd expect based on traditional rankings.
The brands that figure this out early and adjust their content strategy accordingly will have a real advantage. Not because AI Mode is some mysterious black box that requires special tricks, but because most of their competitors are still optimizing for a search experience that's no longer the primary one.
A practical starting point
If you're not sure where to begin, start with observation before action.
Pick 10-15 queries that matter to your business. Run them in Google AI Mode. Note which pages get cited -- yours and competitors'. Look at those pages carefully. What do they have in common? How are they structured? What questions do they answer directly?
That exercise will tell you more about what AI Mode wants than any generic guide (including this one). The pattern you'll likely find: cited pages are specific, direct, and structured for extraction. They answer one thing well rather than many things adequately.
From there, you can prioritize which existing pages to improve and which content gaps to fill. That's a more grounded approach than trying to reverse-engineer an algorithm from the outside.
The underlying principle isn't complicated: AI Mode is trying to build the best possible answer to a complex question by pulling the best available information from multiple sources. If your pages are the best available source for specific sub-questions, they'll get cited. If they're good-but-general, they probably won't.
Traditional rankings measure something real, but they measure a different thing than AI Mode is optimizing for. The sooner your content strategy accounts for both, the better positioned you'll be.