ChatGPT doesn’t search the web for every question. In Semrush’s February 2026 clickstream data, web search was enabled for just 34.5% of observed queries.
That narrows the opportunity for getting a new page cited in a live search response. But it doesn’t make citation strategy irrelevant.
Semrush’s 17-month analysis and Ahrefs’ study of 1.4 million prompts tell us something more useful than “optimize for AI.”
They show how ChatGPT search behavior is changing, which retrieved pages are more likely to get cited, and why the usual SEO assumptions don’t always hold up.
Not every pattern is a ranking factor. And not every correlation gives you something worth optimizing.
Here’s what the data actually says.
TL;DR
- ChatGPT doesn’t search the web for every query. Semrush found web search enabled for 34.5% of queries in its February 2026 US clickstream sample, although usage varies over time.
- ChatGPT retrieves more URLs than it cites. Ahrefs’ analysis of 1.4 million prompts found that approximately half of retrieved URLs received citations.
- Title relevance is associated with citation likelihood. Cited pages showed stronger semantic alignment with both the original prompt and the related fan-out queries ChatGPT generated.
- Reddit retrieval isn’t the same as Reddit citation. In Ahrefs’ dataset, URLs from the dedicated Reddit retrieval category had a citation rate of just 1.93%.
- Prioritize relevance and measurement over AI SEO shortcuts. Improve titles and content around genuine buyer questions, keep pages accessible to search, and measure citations alongside traffic and business outcomes.
Table of contents
How Do ChatGPT and Google Fit Into the Same Search Journey?
ChatGPT and Google can both be part of the same research journey. People may use AI to explore a question, then turn to traditional search for more information or to find a particular website. Semrush’s referral data supports that overlap, although it doesn’t establish why individual users move between the platforms.
The framing that AI is replacing search was always wrong, but now we have numbers to prove it.
ChatGPT referral traffic to the broader web grew 206% in 2025, and the single biggest beneficiary of that traffic was Google:
- Google captures 21% of all outbound clicks from ChatGPT
- The top 10 domains capture 30% of all outbound clicks combined
That doesn’t prove people are discovering brands in ChatGPT and then searching for them on Google. But it does show that these platforms aren’t operating in isolation.
And that’s important for measurement.
If you’re only looking at direct ChatGPT referrals, you could miss some of the influence AI has on the wider buying journey.
Monitor branded search trends, direct traffic, AI referrals, and conversions together. Where possible, use customer surveys and CRM attribution to understand how buyers actually discovered you.
A branded-search increase isn’t proof that ChatGPT caused it. But it’s worth investigating alongside your AI visibility data.
For the measurement side, our guide to LLM traffic versus Google traffic looks at why referral quality and revenue matter more than click volume alone.
How Often Does ChatGPT Search the Web?
In Semrush’s February 2026 US clickstream sample, ChatGPT used web search for 34.5% of observed queries, down from 46% in late 2024. But search usage fluctuated considerably throughout the study, so this isn’t a fixed percentage of all ChatGPT conversations.
Here’s the part that should fundamentally change how you think about “AI search optimization”: ChatGPT only uses web search on 34.5% of queries. That’s down from 46% in 2024.
When ChatGPT doesn’t use live web search, a recently published page has less opportunity to appear as a freshly retrieved source.
But that doesn’t mean two-thirds of all prompts are necessarily answered using training data alone. The model can also work with information supplied in the conversation or through other available context.
And the six-month freshness assumption doesn’t hold up, either. A page being older than six months doesn’t tell you whether ChatGPT will search for it.
What matters is the question being asked, whether the system retrieves external sources, and whether your content is relevant to those sources.
The opportunity is selective. But it’s not governed by a universal rule about how recently a page was published.
What Makes ChatGPT Cite One Page Over Another?
When ChatGPT retrieves sources, it doesn’t cite every URL it encounters. Ahrefs’ research found that cited pages tended to have titles more closely aligned with the original question and its related fan-out queries. That’s a useful pattern, even if it doesn’t reveal ChatGPT’s complete citation-selection process.
In Ahrefs’ dataset, ChatGPT retrieved roughly 33 URLs per prompt and cited approximately half. That doesn’t mean all 33 pages were read in full. Some may have been evaluated using retrieval information such as titles, URLs, and snippets before the model selected which sources to open or cite.
Title relevance was one of the clearest differences Ahrefs observed between cited and non-cited pages.
Freshness was more complicated. The median cited page in the study’s search retrieval category was around 500 days old, but that doesn’t establish that older content performs better generally. And the research wasn’t designed to prove that backlinks don’t matter.
What it does show is that relevance to the question being retrieved is worth taking seriously.
Ahrefs measured semantic similarity between titles and questions using an embedding-based method.
Cited pages averaged 0.602 similarity with the original prompt, compared with 0.484 for non-cited pages. When researchers compared cited titles with their strongest-matching fan-out query, the score rose to 0.656.
These are measurements produced by the researchers, not confidence scores disclosed by ChatGPT.
The pattern suggests that titles matching specific related questions are more likely to appear among cited sources. It doesn’t prove that changing a title alone will produce a citation.
The implication for content teams is fairly straightforward: make the question a page answers obvious from its title and content.
Ahrefs also found a correlation with URL readability. Among search results in the study, URLs with natural-language slugs had an 89.78% citation rate, versus 81.11% for other URLs.
That’s worth considering when creating new pages. But it isn’t evidence that changing an existing URL will improve its citation rate.
And changing URLs creates its own SEO work, including redirects, internal-link updates, and monitoring.
A readable URL is useful. A mass slug rewrite based on one observational study isn’t an optimization strategy.
Does ChatGPT Cite Reddit Content?
Reddit shows up frequently in ChatGPT’s retrieval data. But in Ahrefs’ study, URLs in the dedicated Reddit retrieval category were rarely cited.
Reddit accounted for 67.8% of the non-cited URL pool, while its dedicated retrieval category had a citation rate of just 1.93%.
That doesn’t mean Reddit has no influence on the answer. But retrieval alone can’t prove how the model used the information.
There’s another important distinction in the data. Ahrefs found that 88.46% of URLs in the general search retrieval category were cited. That’s the citation rate within that category, not the share of all ChatGPT citations.
The general search category can also include Reddit and YouTube pages found through ordinary web search.
The takeaway is narrower than “Reddit doesn’t matter.”
Being discussed on Reddit and getting your own website cited are different outcomes. If you’re optimizing for citations, you still need relevant, accessible content that the system can retrieve and reference.
How Can B2B Teams Improve Their ChatGPT Citation Visibility?
Start with content that answers specific buyer questions, titles that describe those answers clearly, and pages that search systems can retrieve. Then compare your citations with competing sources and measure whether the visibility contributes to traffic or business outcomes.
The compounding failure mode here is teams chasing AI visibility metrics that don’t connect to citations, while neglecting the on-page signals that actually drive selection.
A few things that actually matter:
- Write titles that match specific questions. A clear definition, comparison, or implementation question is easier to understand than a vague category label. Use customer research and related searches to find the questions worth answering. Don’t rewrite every title just to imitate a fan-out query.
- Make important pages accessible and descriptive. Use meaningful titles, headings, and URLs. For new content, choose readable slugs. For existing content, don’t change URLs without a clear reason and a proper redirect plan.
- Measure citations alongside actual business outcomes. Track which pages get cited, which competitors appear instead, and whether AI referrals lead to useful visits or conversions. Branded search trends can provide additional context, but they aren’t automatic proof of AI-driven demand.
- Keep the SEO foundations in place. Search retrieval remains an important path into ChatGPT citations. Make sure your pages are indexable, relevant, accurate, and useful. But don’t assume a traditional ranking guarantees an AI citation.
To see how we’ve approached the measurement problem directly, read our AI visibility tracker experiment, where we analyzed citations across 500 Google searches. You can also see how to reverse-engineer YouTube growth with AI and where Google AI Overviews pull their answers from.
What Should You Prioritize to Earn More ChatGPT Citations?
Prioritize existing pages with clear opportunities to answer relevant buyer questions more effectively. Improve their titles, supporting evidence, and content coverage before increasing publishing volume. The research supports relevance as an important factor, but no individual edit guarantees a citation.
The instinct in content marketing is always to publish more. More pages, more keywords, more coverage.
But publishing more pages doesn’t automatically translate into more ChatGPT citations.
Before you add another fifty articles to the calendar, look at the content you already have.
Does each page answer a clear question? Does the title accurately describe that answer? Are you covering the relevant follow-up questions? And are competitors getting cited for information your content doesn’t provide?
That’s a more useful audit than blindly changing URL slugs or chasing another round of AI SEO hacks.
You won’t get cited for every query. ChatGPT doesn’t search for every answer, and retrieval doesn’t guarantee attribution.
The opportunity is in understanding where your content fits, making it worth citing, and measuring what happens when it is.
Frequently asked questions
Why isn’t ChatGPT citing my website?
Your page may not be retrieved for the question, may not match the specific information being sought, or may be retrieved without being selected for citation. Check accessibility, title relevance, content coverage, and which competing sources appear instead.
Does ChatGPT use web search for every answer?
No. Semrush observed web search enabled for 34.5% of queries in its February 2026 US dataset. The rate varies over time and shouldn’t be treated as a fixed percentage for every user or model.
Do page titles affect ChatGPT citations?
Ahrefs found that cited URLs tended to have titles more closely aligned with the original prompt and related fan-out queries. That’s an observed association, not proof that changing a title will independently improve citation frequency.
Should I change my URL slugs to improve AI visibility?
Not solely for that reason. Readable URL slugs correlated with higher citation rates in Ahrefs’ study, but the research didn’t test whether changing an existing URL improves results. Established URLs should be changed only when the benefits justify the migration risks.
How can marketers measure ChatGPT citation performance?
Track brand mentions, cited URLs, citation gaps against competitors, and identifiable referral traffic. Combine those with engagement, conversions, and pipeline reporting. A citation is a visibility event, not proof of a visit or sale.
What to do next
If you’re working on the content side of AI citations, CXL’s Content strategy for LLM visibility and changing search habits course covers researching citation patterns, identifying content gaps, and building content around real buyer questions. For the measurement side, Measuring the modern AI-powered B2B funnel shows how to track AI referrals through GA4, configure reporting, and connect the data to a more complete view of the buyer journey.
Take it further
Finding citation opportunities is useful. Building a process that identifies them, acts on them, and measures the results is better.
The goal isn’t another collection of AI search tactics. It’s a system that helps you see what’s working and decide what to do next.
