How does URL structure relate to AI citations?

ProcrustesPR

Studies have observed a link between URL structure and AI citations: clear, moderately long URL slugs—the final segment of a URL—and natural-language slugs showed better citation performance. However, these findings do not prove that changing a URL will increase citations, nor do they establish an optimal character count across all platforms. A more reasonable approach is to include how clearly a URL describes its content in testing, rather than treat URL structure as a formula that guarantees citations.

ProcrustesPRProcrustesPRWill structured data improve rankings or AI citations?

URL length shows a trend, not a universal threshold

An analysis published by Semrush in 2026 covered 378,000 AI citations. URLs with slugs of 21–25 characters received about 87,000 citations, while those with 6–10 characters received about 57,000, ranking second. The article also reported that slugs of 17–40 characters performed better overall than those of 1–5 characters or 56 characters and above.

These figures describe the study sample: total citation counts are not the same as the probability of an individual URL being cited. The study excerpts do not give a complete account of how many URLs of each length exist in the underlying population. They therefore cannot establish that lengthening a short URL will improve its chances of being cited. The second-place ranking of the 6–10-character group is also a reason not to treat 17–40 characters as a strict requirement.

Citation differences for natural-language URLs depend on retrieval context

In its 2026 analysis, Ahrefs found that, when it restricted source types to search results (search ref_type), URLs with natural-language slugs had a citation rate of 89.78%, compared with 81.11% for those without. This result covers a specific source type and cannot be directly generalized to all AI platforms or all web pages.

The article also found that the titles of cited pages were more closely aligned with the subquestions generated as ChatGPT expanded its queries. However, different source types have different retrieval mechanisms, so combining them in a comparison can be misleading. Taken together, these observations suggest prioritizing the following checks:

  • Whether the slug clearly describes the page, rather than merely meeting a character-count target.
  • Whether the URL, title, and content address the questions in the query.
  • Whether comparisons of citation performance distinguish between source types.

Inferring content from a URL does not mean the page was read

In a 2023 experiment by awoo, ChatGPT was given a nonexistent URL containing keywords for men’s shoes, waterproofing, and Gore-Tex, yet still generated three key points. Based on this, the author suggested that the model might have used the URL structure and keywords to infer the topic and then produce an answer that met the request.

This historical example reminds us that a plausible answer does not prove that a model has accessed a page’s content. However, it was not a study of citation rates and cannot be used to judge ChatGPT’s current browsing capabilities.

Evaluating URL changes also requires care about what the evidence shows: does it support the predicted effect, argue against it, or remain insufficient? A methodological article from Search Engine Journal notes that small samples, short observation periods, or large fluctuations can prevent a test from yielding reliable conclusions. Failing to detect a difference does not prove that no difference exists.

ProcrustesPRProcrustesPRDoes an AI brand mention bring website visitors?

URL structure and AI citations: perspectives from across the web

If an existing URL is very short, should it be lengthened for AI citations?

There is no need to lengthen a URL just because it is short. In Semrush's sample, URLs with slugs of 6–10 characters also ranked second in total citations. The data do not give a complete account of how many URLs of each length exist in the underlying population, so they cannot establish whether lengthening a URL would make it more likely to be cited. If you are considering changes, test clearer descriptions of the content and the removal of unnecessary parameters rather than applying a fixed length.

Why distinguish between source types when comparing URLs' citation performance?

Different source types may have different retrieval mechanisms, and combining them in a comparison allows the sample's composition to influence the results. Ahrefs noted that its analysis had produced aggregate results that were easy to misinterpret when source types were not distinguished. The difference in citation rates for natural-language slugs was observed only after restricting the analysis to search results. That difference therefore cannot be directly applied to other source types.

If a test finds no difference after a URL change, does that mean the change was ineffective?

Not necessarily. A methodological article from Search Engine Journal notes that small samples, short observation periods, large performance fluctuations, or subtle effects can all prevent a test from detecting a difference. Evidence against a predicted effect is not the same as insufficient evidence. You also need to allow time for updated pages to be crawled and indexed, and for systems to respond to the changes.

ProcrustesPRWhat is AI visibility?

References

  1. How Do Technical SEO Factors Impact AI Search? [Study]www.semrush.com · 2026-01-05T15:12:00Z — Discusses: If an existing URL is very short, should it be lengthened for AI citations?
  2. ChatGPT 於 SEO 的兩大錯誤應用—— 工欲善其事,必先避其坑!www.awoo.ai · 2023-05-16T04:57:47+00:00 — Source used in the article above.
  3. Test Like A Scientist: 6 Mistakes That Undermine Your SEO Experimentswww.searchenginejournal.com · 2026-10-13T14:00:21+00:00 — Discusses: If a test finds no difference after a URL change, does that mean the change was ineffective?
  4. Why ChatGPT Cites One Page Over Another (Study of 1.4M Prompts)ahrefs.com · 2026-04-15T14:33:27+00:00 — Discusses: Why distinguish between source types when comparing URLs' citation performance?

Related articles

Will structured data improve rankings or AI citations?Does an AI brand mention bring website visitors?What is AI visibility?How Do SEO and GEO Costs Compare? Align the Scope of Services FirstWhy do publishers need SEO, and what challenges does AI bring?How should you measure search marketing performance and use it to choose partners?Why do publishers still need websites and SEO audits as content discovery changes?How should businesses adjust their search and marketing strategies as search quality, algorithms, and user needs change?
ProcrustesPR
ProcrustesPR