Skip to main content
WhatsApp
Get a free quoteQuote
Digital Marketing

GEO vs SEO for Multilingual Websites: What the Data Shows

A study of 145 million sessions puts all AI assistants combined at under 1% of Google organic traffic, with a higher conversion rate. Here is how to read that for a multilingual site, and how to measure both channels per market.

Michael BastinFounder and Managing Director8 min read
Two colleagues reviewing a localized website on a laptop and a phone while comparing search and AI traffic

Comparing GEO and SEO on Multilingual Sites

Generative engine optimization (GEO) is the work of getting your brand mentioned and cited in answers from AI assistants such as ChatGPT. SEO is the work of earning organic traffic from search engines. If you run a site in several languages, the budget question is simple to state and hard to answer: how much effort should go to each?

One public dataset gives a useful anchor. It does not settle the argument, and it was not built for multilingual sites, but it shows what to measure before you move budget.

What the 145 Million-Session Study Actually Measured

Marie Haynes looked at 12 months of Google Analytics 4 (GA4) data across 100 websites of varying sizes, totaling more than 145 million sessions. She compared sessions arriving from AI assistants with sessions from Google organic search.

Sessions from ChatGPT amounted to just 0.62% of the volume driven by Google Organic. When all other LLMs were included in this number, it increased to 0.73%.Source: Marie Haynes, "GEO Is a Distraction", 12 months of GA4 data across 100 websites

Only 55 of the 100 sites tracked conversions in GA4, so the conversion findings rest on a smaller group. Haynes calls it her small dataset and uses it to argue that chasing ChatGPT citations is optimizing for the wrong target today.

Treat it as one practitioner's analysis, not a peer-reviewed study. The post does not describe the sites' industries, countries or languages, and we found no multilingual breakdown. That gap is the reason this article exists.

Traffic Volume and Conversion Quality Are Different Questions

The same analysis found that AI visitors behave differently from search visitors once they arrive.

Google organic visitors converted at a rate of 3.02%. ChatGPT visitors converted at a rate of 13.8%. [...] Google still drove 96.5% of the total conversions across all channels analyzed.Source: Marie Haynes, "GEO Is a Distraction", 55 sites tracking conversions in GA4

Read the two numbers together. The ChatGPT rate is about 4.6 times the Google rate (13.8 divided by 3.02, our arithmetic), yet Google still produced the overwhelming majority of conversions because its volume is so much larger. Haynes puts the gap at 27 times more conversions from Google organic than from all LLMs combined.

Two cautions apply. A conversion rate calculated on a small number of sessions moves a lot with a handful of leads, so do not expect 13.8% on your own site. And a visitor who asks an assistant a detailed question and then clicks may simply be further along in the decision.

For budget, that suggests a split of roles instead of a contest:

  • Search carries the volume and still pays the bills.
  • AI assistants are a small, high-intent channel worth tracking and nurturing, not yet a replacement for search.

Why Multilingual Websites Need Country-Level Data

An average across 100 sites hides the thing a multilingual business cares about, which is variation by market. Search behavior, assistant adoption and the share of pages available in each language differ from country to country.

A global ratio of 0.73% could mean 0.2% in one market and 3% in another. We do not know, and neither does anyone else without looking at their own numbers.

Three reasons make per-market data worth the effort:

  1. Language coverage. An assistant can only recommend a page that exists in the asker's language. If your German section has 20 pages and your English section has 200, the assistants see two very different brands.
  2. Intent differs by market. A product that people research in a long comparison in one country may be bought on a short search in another.
  3. Your own conversion paths differ. A form, a phone number and a chat widget do not perform the same in every locale.

So before comparing GEO and SEO, split every number by country and by language. The multilingual keyword research guide covers how to find the real terms per market, which feeds both channels.

How to Track AI Traffic Without Trusting Referrals Alone

You will probably need to build your own view of AI assistant traffic in GA4, and you should accept that it will undercount.

Step 1: Create a custom channel group. In GA4 admin, add a channel called "AI assistants" and match session sources against assistant domains. Examples include chatgpt.com, perplexity.ai, gemini.google.com, claude.ai and copilot.microsoft.com. GA4 also has a built-in AI Assistant channel, which Google describes as covering sources like ChatGPT, Gemini, Copilot and Grok, but it does not publish the full source list. Check your own Traffic acquisition report for the exact source names your visitors arrive from, because they change.

Step 2: Segment by country and landing page. Add Country and Landing page as dimensions, then compare sessions, engaged sessions and key events for AI assistants against Google organic in each market.

Step 3: Plan for missing referrers. Some visits from assistants arrive without referrer information and appear as direct traffic. When direct traffic to deep pages, such as a specific service page in Italian, rises with no campaign behind it, note it as a possible AI effect and do not count it as proof.

Step 4: Add a "how did you hear about us" field. A free-text or dropdown field on your contact and quote forms lets people say "ChatGPT" or "an AI assistant" in their own words. It is crude, and it catches the visits referrers miss.

Step 5: Test your own visibility directly. Once a month, ask the assistants a fixed list of questions in each target language and record whether your brand is mentioned and which page is cited. Treat the results as a log of samples, since answers vary between runs.

Where GEO Fits Into an International SEO Strategy

GEO is not a separate discipline with its own foundations. The work that makes a page easy for an assistant to quote overlaps heavily with good international SEO: clear structure, accurate information, pages that exist in the reader's language and a site crawlers can reach.

If you need the technical groundwork, our international SEO guide covers site structure, hreflang and localization. This article takes a different angle: how to divide effort and prove what works. If you are weighing a vendor for either channel, our guide to choosing a multilingual SEO agency lists twelve questions, including how each one handles AI search.

A practical division looks like this:

  • Foundation (SEO). Indexable pages in every target language, correct hreflang, native keyword research, fast pages. This serves both channels.
  • Clarity (shared). Direct answers near the top of key pages, specific facts, named entities and consistent product naming across languages.
  • Reputation (GEO). Mentions on sources assistants are likely to draw from, such as industry publications and review sites, in each language you sell in.
  • Measurement (both). The GA4 view above, plus Search Console by country.

None of this requires abandoning search. If your localized pages do not rank, they are also unlikely to be recommended, so fix coverage and quality first, then layer on the reputation work.

A 90-Day SEO and AI Visibility Measurement Plan

Use the first quarter to collect evidence, not to reorganize your budget.

Days 1 to 30: Baseline.

  • Build the AI assistants channel group in GA4 and confirm conversions are tracked as key events. Remember that 45 of the 100 sites in the study were not tracking any, so the numbers we are discussing exist only because most sites had set up tracking.
  • Export Google organic sessions, key events and conversion rate by country and language.
  • Record a fixed set of 20 to 30 buyer questions per language, and log whether assistants mention you.

Days 31 to 60: Fix and publish.

  • Close the biggest coverage gaps: priority pages that exist in your main language but not in a market where you sell.
  • Rewrite weak answers on those pages so the first paragraph states what you offer, for whom and where.
  • Add the "how did you hear about us" field if you have not already.

Days 61 to 90: Compare and decide.

  • Repeat the visibility test and compare it with your baseline.
  • Compare AI assistant sessions and conversions with Google organic by market.
  • Decide where to invest next, using a simple rule: if a market's AI sessions and conversions are rising and your search position is already strong, shift some effort toward reputation work there. If search is weak, repair search first.

SEO-GEO Checklist for Multilingual Businesses

Use this list as a quarterly review.

  • Every priority page exists in each target language and is indexable.
  • Hreflang annotations are correct and reciprocal.
  • Keyword research was done in each language, not translated from English.
  • GA4 key events are tracked, and conversions are visible by country.
  • A custom "AI assistants" channel group exists and is reviewed monthly.
  • Contact and quote forms ask how the visitor found you.
  • Key pages open with a direct, specific answer in the page's own language.
  • Product, brand and service names are used consistently across languages.
  • A fixed question set is tested in each language every month, with results logged.
  • Budget decisions cite per-market data, not global averages.

What to Do Next

The data supports a measured position: keep search as the base of the strategy, track AI assistants as a small channel with strong conversion, and let your own per-market numbers decide how far to go. A multilingual site has more to lose from guessing than most, because one average can hide very different markets.

If you want a second pair of eyes, request a free multilingual SEO audit. We will compare Google, AI assistants and conversions per market, and show where your language versions are strong or missing. You can also see how our multilingual SEO services fit around a professional translation workflow carried out by human translators, with AI post-editing available on request.

FAQ

Is GEO replacing SEO for multilingual websites?

Not on this evidence. In the study discussed above, all AI assistants combined sent under 1% of the sessions that Google organic did. Search still drives most traffic and most conversions, while AI referrals are small and worth tracking.

Why do AI visitors convert at a higher rate?

The study reports the difference but not the cause. A plausible explanation is that people who click through from an assistant have already narrowed their options, but that is an inference and your own data may differ.

How do I find AI traffic in GA4?

Create a custom channel group that matches the source names of assistants, review it by country, and add a self-reported source question to your forms. Expect some AI-driven visits to appear as direct traffic.

Share this article

4.96/5.0 from 178 Google reviews

Ready to Reach Readers in Every Language?

Tell us which markets you are targeting and we will show you how translation and multilingual content work together to reach them.

  • 20+ Years experience
  • 500+ Native experts
  • 50+ Language pairs
Open WhatsApp chat