URL Slugs in the Age of AI Search

For about twenty years, "URL slug SEO" meant Google. Slugs that ranked well in Google were the slugs that mattered — and the playbook was stable: hyphens, keywords, under sixty characters, no stop words.

That playbook is still correct. But it's no longer the whole story. ChatGPT, Perplexity, Claude, and Gemini are now serving billions of search-like queries every month, and they all factor URLs into their citation decisions in ways that look familiar but behave differently. Pages that look fine to Google can be invisible to AI search if their slugs don't communicate clearly.

This is what changed, why it matters, and what to do about it.

The shift: AI search engines cite URLs as sources

When ChatGPT or Perplexity answers a question by citing web sources, it picks a small handful of pages — usually three to ten — from a much larger candidate pool surfaced by an internal search index. The selection is based on a mix of signals: domain authority, content depth, freshness, and (this is the relevant part) URL legibility.

An AI engine has milliseconds to decide which pages to cite. It doesn't fully render every candidate page. It looks at high-signal metadata: the page title, the meta description, and a sampled section of the body — and the URL itself, parsed for topic clarity.

If the URL clearly states the page topic, the engine has higher confidence the page is relevant. If the URL is opaque (numeric IDs, encoded parameters, fragments), the engine has lower confidence. Lower confidence means lower citation probability.

What ChatGPT and Perplexity see when they pick citations

An AI engine's URL parser is closer to Google's pre-2015 model — pattern-matching and keyword extraction — than to Google's modern intent-based ranking. AI engines are, in a sense, less forgiving of bad URLs because they have less computational budget per candidate to compensate for opacity.

For a query like "how to write a markdown table," consider these two URLs:

• example.com/post-4729
• example.com/blog/how-to-make-a-markdown-table

Both pages might have identical content. Google ranks them similarly — the algorithm looks at body content, links, and behavior. AI search engines don't always go that deep. The second URL is more likely to be cited because the URL itself confirms the topic match.

Why a bad slug loses you AI citations even if you rank in Google

Here's the asymmetry that catches publishers off-guard: you can rank #5 in Google search results and never get cited by Perplexity for the same query. Different signal weights. Different evaluation pipelines.

The publishers being cited heavily by AI engines in 2026 share three URL traits:

1. The slug names the topic in plain English.
2. No nested category paths burying the slug.
3. No date or numeric ID in the URL.

This is essentially the same advice as Google SEO, with sharper edges. If Google rewards a clean URL with a 5–10% lift, AI search rewards it with closer to a 30–40% lift in citation probability — because the URL is doing more of the relevance work.

The 4 slug patterns AI engines reward

1. Topic-first slugs

Lead with the head term. /markdown-tables-guide, not /guide-to-markdown-tables. The first words in the slug carry the most weight in the AI engine's relevance scoring.

2. Question-style slugs for answer-engine queries

For pages targeting how-to or what-is queries, slugs that mirror the question pattern get cited more. /how-to-extract-emails-from-text beats /email-extraction-guide for AI engines, because users ask the question literally and the engine matches token-by-token.

3. Year-tagged slugs for time-sensitive content

For content that changes year-to-year (technology, regulations, best practices), include the year. /markdown-guide-2026 signals freshness in a way that AI engines can read directly. Refresh the date when you update the content.

4. Single-segment slugs over deep paths

AI engines parse URL paths token-by-token. /blog/seo/url-slugs/2026/the-definitive-guide dilutes the topic across five segments. /blog/url-slug-seo-the-definitive-guide-2026 concentrates it. Flat is better.

Real examples — same article, three slugs, three AI outcomes

An anonymized publishing test, late 2025, three identical articles published on three different domains with three different slugs:

Google ranking barely differed across the three. Perplexity citation differed by an order of magnitude. The URL was carrying the topic signal in a way that mattered to the AI engine but not to Google's deeper-evaluation pipeline.

Action plan: audit your existing URLs before year-end

Take your top 20 highest-traffic pages. For each, ask:

• Does the slug clearly state the topic without external context?
• Is the head keyword in the first segment of the slug?
• Are there nested categories burying the slug?
• Does the slug have a numeric ID, date, or opaque fragment that adds no information?

If any answer is "no" for a page that's getting AI search traffic, you have a candidate for an SEO-aware slug rewrite. Don't rewrite blindly — every change requires a 301 redirect, and high-traffic pages need careful migration. But for new content going forward, the AI-search-aware slug is essentially free upside.

What's coming next

AI search engines will continue to tighten URL parsing, not loosen it. The trend over the past two years has been toward more strict topic matching, not less — engines have learned that publishers will optimize whatever signal they're using, so they keep the bar high. Slug quality will keep mattering as long as URLs are part of the citation decision pipeline.

For the underlying SEO theory and the seven rules of slug optimization, see URL Slug SEO: The Definitive Guide for 2026. For the production workflow that ships AI-clean output, see format ChatGPT/Claude output for production.