AI search traffic by page type: where LLM referrals land

80%+

of AI referral traffic hits just three page types

36.45%

goes to free tools alone

3.6%

lands on pages that do not exist

THE HEADLINE FINDINGWhen Ahrefs traced its own LLM referral traffic back to the pages receiving it, free tools, product pages, and the homepage absorbed more than four fifths of everything. The blog, the surface most teams spend their entire budget on, was a rounding error by comparison.

TL;DR · 60 SECONDSAI search traffic by page type is far more concentrated than organic traffic. Free tools took 36.45%, product pages 23.12%, and the homepage 20.41% in the Ahrefs breakdown. That is a different distribution than classic search, and it means the standard GEO playbook of publishing more articles is aimed at the wrong pages. Fix your product and tool pages first.

Almost every generative engine optimization program starts in the same place: publish more content, structure it for extraction, and wait for the citations. That advice is not wrong, but it quietly assumes AI engines send people to articles. The best public breakdown of AI search traffic by page type says otherwise.

Ahrefs' Patrick Stox pulled the referral traffic that large language models sent to ahrefs.com and grouped it by the kind of page it landed on. The result was lopsided in a way organic search never is. Three page types took more than 80% of the total, and the two that dominate most content calendars, blog posts and long-form guides, were not carrying the load. You can read the full breakdown in Ahrefs' analysis of LLM referral traffic.

One site is one site, and we will get to the limits of that sample. But the shape of the finding matches what we see instrumenting generative engine optimization programs for enterprise clients: the pages that get named in an answer and the pages that actually receive the click are frequently not the same page, and the gap is widest for companies whose entire AI strategy is a blog.

Where AI search traffic by page type actually lands

Here is the distribution Ahrefs reported. Read it as a ranking of what AI engines consider a useful destination, not as a universal constant. The absolute percentages will differ for your site. The concentration almost certainly will not.

PAGE TYPE	SHARE OF AI REFERRAL TRAFFIC	WHAT IT TELLS YOU
Free tools	36.45%	Engines route people to something they can immediately use
Product pages	23.12%	The answer names a product, the link goes to the product
Homepage	20.41%	Brand-level queries resolve to the front door
Everything else	~20%	Blog, guides, docs, and international pages combined

The tail is the part worth sitting with. Every blog post, every pillar guide, every comparison page and every localized version of the site is competing for what is left after the top three take their cut. That is a brutal ratio if your content plan assumes articles are the destination.

THE DISTINCTION THAT MATTERSCitations and clicks are two different metrics with two different winners. An article can earn the citation that makes the engine confident enough to recommend you, while the click lands on your homepage or your pricing page. Measuring only one of them will make your program look either better or worse than it is.

Why free tools and product pages beat the blog

This is not an accident of one company's site architecture. Three mechanics push AI referral traffic toward utility pages, and all three are structural.

1The engine already answered the questionIf someone asks how to check a backlink profile, the model explains the concept in the answer itself. What it cannot do inside the chat is run the check. So the link it offers is the tool, not the article explaining the tool. The informational job is done before the click exists.

2Product pages are unambiguous entitiesModels resolve queries to entities. A product page states what the thing is, what it costs, and who it is for, in a form that maps cleanly onto a recommendation. A blog post about the category does not identify a purchasable thing, so it feeds the answer without earning the destination link.

3Brand queries collapse to the homepageA meaningful share of AI prompts are navigational in disguise. Someone asks what a company does or whether it serves their segment, and the engine sends them to the front door. That traffic is high intent and almost entirely invisible in a content-first reporting setup.

Semrush's analysis of AI search traffic points the same direction from a different angle: in their study, 50% of the links included in ChatGPT responses pointed at business and service websites rather than publishers, and the average AI search visitor was worth 4.4x the average traditional organic visitor on conversion rate. Fewer visits, better visits, landing on commercial pages. We unpacked the economics of that ratio in our piece on why AI search converts at a different rate.

Free tools36%

Product pages23%

Homepage20%

All other pages21%

Share of AI referral traffic by page type (Ahrefs, first-party analytics)

Notice that the same Semrush work found ChatGPT citing pages that rank in position 21 or worse for the related query almost 90% of the time. Retrieval is not a mirror of the rankings. A page can be invisible in classic search and still be the source an engine leans on, which is exactly why which signals actually earn a citation is a separate discipline from ranking work.

The international blind spot nobody is measuring

The most under-discussed number in the Ahrefs breakdown has nothing to do with page format. International pages took 3.1% of AI search traffic, against 22.2% of organic search traffic. That is a seven-fold collapse, and it is the single largest disparity in the dataset.

22.2%

of organic traffic goes to international pages

3.1%

of AI search traffic does

the gap between the two channels

If you run a localized site, this is the finding to act on. Whatever the cause, and thin localization, weaker link equity to translated URLs, and English-dominant training and retrieval are all plausible contributors, the practical effect is the same: your non-English pages are earning a fraction of the AI visibility their organic performance would predict. Enterprise teams with twenty locales are effectively running twenty untested experiments here.

The other oddity in the same data: 3.6% of AI referral traffic arrived at pages that do not exist, mostly from ChatGPT inventing plausible URLs. That is a real and recoverable loss. Those are people who wanted your page badly enough to click a link to it. A 404 log filtered to AI referrers is one of the cheapest wins available, and the fix is usually a redirect, not a rewrite.

DO THIS TODAYFilter your 404 report to referrals from ChatGPT, Perplexity, and Claude. Every hallucinated URL with repeat hits is a page an engine believes you should have. Redirect the close matches. Build the ones that get requested repeatedly.

What AI search traffic by page type means for your roadmap

The honest read on this data is not that content does not matter. Content is what makes the engine confident enough to name you. The read is that content and destination are different jobs, and most programs fund one of them. Here is how we sequence it.

Week 1

Audit your product pages as answer targetsEvery product page should state, in extractable prose near the top, what the product is, who it serves, what it costs or how pricing works, and what it replaces. Most enterprise product pages open with a positioning slogan and bury all four. That is a page an engine cannot quote.

Quarter

Ship one genuinely useful free toolThe single biggest category in the data is the one most B2B sites do not have. A calculator, checker, or generator tied to your category gives engines a destination that answers the part of the question chat cannot. It does not need to be sophisticated. It needs to work without a signup.

Week 2

Treat the homepage as a citation surfaceBrand-level prompts resolve here. Your homepage should carry a plain-language description of what you do and who you serve, not just a hero line. This is the page most likely to be quoted verbatim in an answer about your company.

Ongoing

Keep publishing, but change what you measureArticles earn the citation that produces the recommendation. Judge them on mention rate and citation rank, not on referral sessions. Judging article performance by AI referral clicks will tell you to stop doing the thing that is working.

That last point is where we see the most expensive mistakes. A team pulls an AI referral report, sees that the blog drove 4% of AI sessions, and cuts the content budget. Six weeks later mention rate falls across the board, because the articles were the corroboration the engines were using to recommend the product pages in the first place. The click and the cause live on different URLs.

“Your articles buy the recommendation. Your product pages collect it. Cut the first to fund the second and you lose both.”

How to instrument this in your own analytics

None of this is worth arguing about in the abstract when you can measure it on your own domain in an afternoon. AI referrers pass identifiable hostnames. Segment on them, then group the landing pages by type.

AI referrer patterns worth segmenting● LIVE

# Referrer hostnames that indicate an AI-assistant click
chatgpt.com
chat.openai.com
perplexity.ai
claude.ai
copilot.microsoft.com
gemini.google.com
 
# Group landing pages into buckets, then compare against organic
/tools/*        -> free tools
/products/*     -> product pages
/              -> homepage
/blog/*         -> editorial
/[locale]/*     -> international
 
# Two reports that matter more than sessions
1. AI landing pages by type, indexed against the same split for organic
2. 404s where the referrer is an AI assistant (hallucinated URLs)

Run that split against the same breakdown for organic search. The comparison is the insight, not the raw AI number. If your product pages take 8% of organic traffic and 30% of AI traffic, you have just found where to spend. This is the reporting build we put in place at the start of most reporting and analytics work, because without it every conversation about AI visibility is anecdote.

One caveat worth stating plainly, because the number is going to get quoted without it: the Ahrefs figures come from a single site with an unusual profile. Ahrefs has a large free tool suite and a category-defining brand, which inflates the tool and homepage buckets relative to a typical B2B company. Ahrefs' own write-up notes the tracking is incomplete and that AI Overviews traffic is lumped in with organic. Treat the ranking as directional and the exact percentages as theirs, not yours.

Where to start this week

Pick your ten highest-intent prompts, the ones a buyer would actually type before a purchase. Run them across ChatGPT, Perplexity, and Google AI Mode. Write down two things for each: whether you are cited, and which URL the engine offers when it recommends you. That second column is the one nobody records, and it is the one that tells you which page to fix.

Then compare it against the traffic split above. If the engines are naming your articles but linking your homepage, your homepage copy is now your most important content asset. If they are not naming you at all, the problem is upstream of page type entirely, and the work starts with the extractable structure and corroboration that make a brand quotable. We walked a technical category through exactly that sequence in a category build for a technical buyer, and the same rule held: the format that earns the mention and the page that earns the click were never the same file.

Do that audit before you commission another twenty articles. The content is probably not the problem. The destination is. And if you want the format-level view of what earns the mention in the first place, comparison pages still earn the most citations by a wide margin, which is a useful reminder that the article is doing its job even when it is not the page collecting the visit.

See where you are cited today

A free snapshot audit of your rankings and AI citations before we ever talk.

Schedule a free consultation

Josh BernsteinMANAGING PARTNER, SOMETHING INC.

Josh leads work at the intersection of SEO and generative engines at Something Inc., helping B2B brands get ranked and cited across every major AI engine.

AI search traffic by page type: your blog is not the target