GEO

Measuring AI referral traffic in Saudi Arabia

How to find, tag and track the visits ChatGPT, Gemini and Perplexity actually send a Saudi business, and why most of it hides as direct traffic.

By Jeremy ZavaAI-assisted, reviewed by Ibrahem HelalPublished 29 September 20266 min read

AI referral traffic is the visits a website receives after ChatGPT, Gemini or a similar assistant names and links it in an answer, and most analytics setups undercount it because a large share arrives as unattributed direct traffic instead of a labelled referral. Watching it needs a deliberate setup, not the default dashboard.

This is the measurement half of how Saudis use AI search to find businesses: once a brand starts getting cited, this is how you find out whether anyone actually clicked through.

On this page

What actually counts as AI referral traffic?

A visit where the previous step in the journey was a conversation with an assistant that named and linked your page, rather than a search results click or a direct type-in. It sits in a strange middle ground: caused by a citation, but arriving to your analytics looking like almost anything else.

That ambiguity is the whole difficulty of measuring it. A Google click carries a clear referrer. A social share carries a clear referrer. An assistant conversation, depending on the app and the platform, frequently carries none at all, which pushes the visit into whatever bucket your analytics tool uses for traffic it cannot explain.

Worth separating from this, too, is traffic from the crawlers themselves rather than from a person reading an answer. OpenAI, for instance, publishes separate crawlers for training, for fetching a page live during a conversation, and for building the index behind citations [3]. None of those is a "visit" in the sense this article means; they are automated fetches, and a spike in bot traffic from one of them tells you your page is being read, not that a person clicked through from an answer.

Which sources should you actually watch for?

A specific, growing list of referrer domains, plus the general pattern of what those visits look like once they land.

These show up in the standard Traffic acquisition report once you know to filter for them, since the report groups traffic by session source and medium across the whole site [1]. The visible number from this filter is real, but it is the floor, not the ceiling, of what is actually happening.

The list itself is worth revisiting every few months rather than treated as fixed. New assistants launch, existing ones change their domain or add a separate mobile-app referrer, and a filter built once and forgotten quietly stops catching a growing share of the traffic it was built to find. Anyone running this check regularly should keep the domain list itself under review, not just the numbers behind it.

Why does so much of it show up as "direct" traffic instead?

Because the referrer header, the piece of information a browser normally sends saying "this visit came from that page," is often stripped or never set by assistant apps, particularly on mobile where a lot of Saudi AI use happens. Analytics tools built around that header have nothing to attribute the session to, so it defaults to direct or unattributed.

This is not a Saudi-specific quirk; it happens everywhere AI apps are used on mobile. It matters more here specifically because of how mobile-heavy Saudi internet use already is, which means whatever share of AI traffic hides this way elsewhere is likely hiding at an even higher rate in this market.

There is a second, quieter cause worth knowing about: some in-app browsers used by assistant apps strip or rewrite the referrer for privacy reasons before the page ever loads, regardless of platform. That is a deliberate design choice by the app, not a bug in your analytics, and no tagging change on your end will fully undo it. The realistic goal is catching more of the visible signal, not eliminating the gap entirely.

How do you set up tracking that catches more of it?

With a short list of habits added on top of the default analytics setup, not a full rebuild.

What does healthy AI referral traffic actually look like?

Small, for now, and disproportionately valuable per visit. The volume is genuinely a minority of total traffic for almost every business today, Saudi Arabia included, but a visitor arriving after an assistant already recommended you specifically has, in effect, already been through a comparison you did not have to run yourself.

We tell clients not to judge this channel by volume yet. Judge it by what the visitor does once they land, because someone arriving off the back of an assistant's recommendation converts differently than someone who clicked a tenth blue link out of curiosity.

That is worth checking directly rather than assuming: compare bounce rate and conversion rate for the segment of sessions from known AI referrer domains against your site average. Even a small sample often shows a meaningfully different pattern.

Saudi Arabia's digital audience is large enough that even a currently small share of it translates into a real number of people. With 34.4 million people online at the end of 2025, a fraction of one percent of that audience touching an AI assistant before finding your business is still a population worth building a measurement habit around now, rather than waiting until the dashboard number looks impressive on its own [4].

How does this connect back to whether your brand gets cited at all?

Directly, and in sequence. Referral traffic is downstream of citation; a brand invisible to assistants has no referral traffic to measure regardless of how well the analytics are set up. Which Saudi brands AI already cites is the citation-side check that should come first, and Gemini in Saudi Arabia covers the one assistant most tightly wired into the Google ecosystem this kind of traffic often comes through.

The broader context for why this channel is worth building measurement for now, even while it is small, sits in AI adoption in Saudi Arabia: the audience asking these questions is large and growing quickly, which means the traffic behind it will not stay this small for long.

What should you actually do with this data each month?

Track it, do not act on a single week of it. Save the referrer segment, check it alongside your citation audit, and write one sentence each month about what moved and why, the same discipline that makes the citation tracking useful rather than just a snapshot.

Over two or three quarters, that habit produces the only real record of whether AI visibility work is paying off in visits, not just in mentions. Anyone selling AI search work who cannot show you this kind of trend, rather than a single flattering screenshot, is not measuring it properly either. An AI visibility audit is the place to set up both halves of this at once.

Keep the write-up short and specific: which referrer domains moved, what the citation check found that same month, and one sentence on whatever you changed in between. Six months of those short notes, read back to back, tell you more about whether this channel is working than any single month's dashboard ever will, and they are the record you will actually want when deciding whether to expand the effort next year.

Questions people also ask

• Does Google Analytics show AI referral traffic automatically?

  It shows some of it, as referral traffic from domains like chatgpt.com or perplexity.ai, in the Traffic acquisition report. It misses the share that arrives as unattributed direct traffic, which happens often enough that the visible number understates the real one.

• Why does AI traffic often show up as 'direct' instead of 'referral'?

  Because many assistant apps, especially on mobile, do not pass a standard referrer header the way a browser link click does. Analytics tools that rely on that header have nothing to attribute the visit to, so it defaults to direct.

• Is a small amount of AI referral traffic today a sign it isn't worth the effort?

  Not on its own. The volume is genuinely small for most businesses right now. What matters more is the trend line and the quality of the visits, since someone arriving after an assistant's recommendation is often closer to a decision than an average visitor.

• How often should I check this?

  Monthly is enough to see a trend without overreacting to noise. Weekly if you have recently made a citation-focused change and want to see whether it moved anything.

Sources

1. [GA4] Traffic acquisition report — Google Analytics Help (checked 23 September 2026)
2. I tracked 13,184 AI citations across 4 LLMs: here's what I learned — Orbit Media Studios (checked 23 September 2026)
3. Overview of OpenAI crawlers — OpenAI (checked 23 September 2026)
4. Digital 2026: Saudi Arabia — DataReportal (Kepios) (checked 23 September 2026)

Set up citation and referral tracking together

Book a free 30-minute consultation with Ibrahem and Jeremy. We look at your site while we talk, and you leave with a prioritised plan.

Book a free 30-minute consultation

Who wrote this

Jeremy ZavaCo-founder

Leads growth and client partnerships.

Works on: CRO, SEO

Reviewed before publishing by Ibrahem Helal, Co-founder.

Posts by Jeremy Zava

More on this

• Which Saudi brands do AI assistants cite?A method for checking whether ChatGPT and Gemini cite your brand for real buying questions, with a worked hypothetical example of what the gap looks like.
• Gemini in Saudi Arabia: what it isWhat Google Gemini is, how it differs from a Google search, whether it works well in Arabic, and how a Saudi business gets cited by it.
• AI search vs Google in Saudi ArabiaHow asking ChatGPT or Gemini differs from Googling in Saudi Arabia today, how much search still goes to Google, and what each means for getting found.
• GEO for hotels in Saudi ArabiaHow Saudi hotels and hospitality brands get named when a traveller asks ChatGPT or Gemini where to stay, and what actually feeds that answer.