AI Overviews Don't Answer in French Yet | Canada GEO

Ask an English-Canadian buyer a purchase question on Google and there is a fair chance the AI Overview answers it directly, right inside the results page, before a single blue link is clicked. Ask the same question in French in Montreal, and as of this research, Google's AI Overview does not answer at all. Google Canada's own blog states plainly that French is "not currently supported" for AI Overviews, with no confirmed launch date. That single, unresolved gap changes how AI-search visibility has to be built for any brand serious about Quebec.

This matters more than it might first appear. Canadian business AI adoption reached 19.2% in Q2 2026, up from 12.2% just two years earlier, according to Statistics Canada's Canadian Survey on Business Conditions. On the consumer side, 56% of Canadians surveyed say they have tried ChatGPT, a trial rate that actually exceeds the 45% reported in the same study for the United States. Canadians are not lagging in AI adoption. What's different is which AI surface they land on, and that difference runs almost entirely along the English/French line.

The Gap, Specifically

The precise timeline matters here, because "AI Overviews" and "AI Mode" are two different Google features that rolled out on two different schedules, and conflating them produces an inaccurate picture. AI Overviews, the answer box embedded directly in regular search results, launched fully in Canada on October 28, 2024, in English, Hindi, Indonesian, Japanese, Portuguese and Spanish, with French explicitly excluded at that launch. AI Mode, a separate, more conversational full-page experience accessed through its own tab, launched in Canada on August 21, 2025, also English-only at first. French support for AI Mode specifically arrived later, expanding globally to French and more than 35 other languages on October 6, 2025.

What this means in practice: AI Mode, the newer, more chat-like feature, now does support French, including in Canada, as of October 2025. AI Overviews, the more prevalent feature embedded directly in ordinary search results, is the one Google Canada's own blog described as not currently supporting French, with no confirmed date given for when that might change. For a Quebec-based searcher typing an ordinary question into Google, the AI Overview box most searchers actually encounter still doesn't answer in French, even though a searcher who deliberately switches to the AI Mode tab now can. That's a narrower, more precise gap than "Google doesn't do French AI search in Canada," but it's still a meaningful one, because AI Overviews is the default, higher-visibility surface and AI Mode requires a searcher to actively choose it.

Google has been closing some of this distance from other directions too. Search Live, a voice-forward AI search feature powered by Gemini, expanded to Canada in both English and French as of March 2026. Taken together, the pattern looks less like a single permanent gap and more like a staggered rollout where French support arrives feature by feature, sometimes many months after the English equivalent. That staggering is itself useful information for planning: it suggests the AI Overview gap specifically is more likely a matter of when than if, which is exactly why a bilingual GEO strategy needs a recurring re-test built in rather than a one-time assessment.

None of this changes the underlying opportunity in the meantime. Canadian business AI adoption reached 19.2% in Q2 2026, up from 12.2% just two years earlier, according to Statistics Canada's Canadian Survey on Business Conditions. On the consumer side, 56% of Canadians surveyed say they have tried ChatGPT, a trial rate that actually exceeds the 45% reported in the same study for the United States. Canadians are not lagging in AI adoption. What's different is which AI surface they land on for a French-language query today, and that difference still runs largely along the English/French line even as some of Google's own features begin to close it.

Data from NETendances 2025, an annual survey run by Université Laval, shows that 52% of Quebec internet users had used a generative AI tool by October 2025, up from 33% a year earlier, with adoption rising fastest among users under 55. Among Quebec's AI users specifically, the engine mix looks like this: ChatGPT leads with 84% usage, Microsoft Copilot follows at 29%, and the standalone Gemini app sits at 22%. A meaningful share of that traffic, at least for AI Overview-style queries specifically, still doesn't touch Google's in-search AI layer for French-language questions, because that particular feature isn't fully there yet to catch it.

There's a deeper, structural reason this gap isn't just a rollout scheduling issue. Most AI models are trained predominantly on English-language data, with French supplemented primarily from European, Parisian-French sources rather than Quebec French ones. Quebec French differs from European French in vocabulary, idiom and cultural reference in ways that matter for retrieval: words like "magasinage" (shopping) or "courriel" (email) have no exact European French equivalent and can be mishandled by a model trained mostly on a European French corpus. The Canada School of Public Service has noted explicitly that a tool like ChatGPT "may not provide the appropriate regional language variant, such as Canadian French, and it likely won't warn you about this." The Berkeley School of Public Policy's research on Quebec AI equity treats this as a structural risk of digital marginalisation, not a minor edge case. There's also a documented behavioural quirk worth knowing: in at least one enterprise implementation, simply specifying a Montreal, Quebec location in an AI system's context caused it to respond in French even when the original question was asked in English, which means location signals alone can shift which language, and therefore which source layer, an AI system draws its answer from.

Why This Isn't Just a Translation Problem

The instinct for a lot of brands is to treat this as a translation task: take the English page, run it through a French version, done. That instinct misses two separate problems. The first is legal. Quebec's Bill 96, an amendment to the Charter of the French Language in force since June 1, 2025, requires any commercial website reaching Quebec residents to present French content with prominence at least equal to English, enforced by the Office québécois de la langue française. The OQLF received 9,125 complaints in the 2023-2024 period, a 33% year-over-year increase, which signals genuinely active enforcement rather than a rule sitting quietly on the books.

The second problem is that machine-translated content consistently underperforms genuinely localised Quebec French, both in organic search and in the AI-citation testing this research draws on. A model resolving a French query doesn't just need French words, it needs the vocabulary, phrasing and cultural reference points that a Quebec-based content reviewer would actually use, not a literal rendering of an English original. Content built this way earns citations across ChatGPT, Copilot and Gemini in a way a translated page structurally cannot.

What a Bilingual GEO Strategy Actually Requires

Given the surface split, a workable approach separates the two language tracks almost entirely rather than treating French as an English page's shadow.

• Separate content architecture. en-CA and fr-CA sections built and reviewed independently, not templated from one to the other, with correctly implemented hreflang tags connecting them.
• Engine-specific weighting. English content optimisation can lean on Google's AI Overview behaviour. French content optimisation has to prioritise ChatGPT first (84% share), then Copilot and Gemini, since Google's AI surface isn't in the mix yet.
• Native-speaker review as a gate, not a polish step. Quebec French content should be written or reviewed by an in-market speaker before publication, not translated and lightly checked afterward.
• Re-verification on a schedule. Google's own blog language ("hope to have an update soon") signals this could change without much notice. A strategy built assuming the gap is permanent needs a trigger to re-test once French AI Overview support does launch.

Signal	English Canada	Quebec French
Primary AI surface	Google AI Overviews / AI Mode	ChatGPT, Copilot, Gemini (no Google AI surface)
Dominant engine share	Not separately measured in this research	ChatGPT 84%, Copilot 29%, Gemini 22% (Quebec AI users)
Generative AI trial rate	56% have tried ChatGPT (national figure)	52% had used generative AI by Oct 2025, up from 33% a year earlier
Legal content requirement	None specific to language prominence	Bill 96: French at least as prominent as English

The Other Half of the Problem: Canadian Sources Are Under-Cited Too

The language gap isn't the only reason Canadian content struggles to get cited by AI engines. There's a separate, compounding issue: Canadian domains are systematically under-represented in the training data most AI models draw on, meaning even solid English-Canadian content competes at a structural disadvantage against the much larger pool of US content covering similar topics. One documented Canadian citation study found that a single directory, Threebestrated.ca, was cited 116 times across a dataset of 1,732 citations, more than any individual business in that dataset, a strong signal that AI engines currently rely heavily on a small number of established .ca directories to establish Canadian relevance, rather than deeply indexing individual Canadian business websites.

There's a broader tailwind worth factoring into content strategy here too. The Canadian Internet Registration Authority (CIRA) reports 3.43 million .ca domains under management, growing faster than the global country-code average, with a notable surge in late 2024 tied to "buy Canadian" consumer sentiment. That pattern suggests AI models are increasingly weighting Canadian-entity signals, .ca domains, Canadian addresses, CAD pricing, references to Canadian regulators, as relevance indicators in their own right, somewhat independent of raw content volume. For a GEO strategy, the practical implication is that citation-building work should prioritise .ca directory placement, Canadian review aggregator management, and complete, consistent Google Business Profile data as foundational infrastructure, before content production even begins, rather than treating directory presence as an afterthought to a content calendar.

The First-Mover Window

No identified Canadian agency currently publishes a dedicated French-Canadian GEO methodology. That's a genuine gap, not a marketing line: most Canadian SEO and GEO providers are concentrated in Toronto, Vancouver, Calgary and Ottawa, with English-first delivery models, and Quebec-specific specialists tend to focus on traditional SEO or paid media rather than AI-citation work. A brand willing to build genuine Quebec French content now, before Google's AI Overview gap closes, is building citation history in ChatGPT, Copilot and Gemini's French-language indexes while most competitors are still treating French as an afterthought.

None of this means English-Canadian GEO work should slow down. AI Overviews already touch a large and growing share of Canadian commercial queries, and English content optimised for Google's AI surface reaches the larger population by raw numbers. The point is narrower: a brand that stops at English is leaving Quebec's fastest-growing AI-user segment to competitors, or more often, to nobody at all, because the market genuinely hasn't been built for yet.

---

Frequently Asked Questions

Does Google's AI Overview work in French anywhere in Canada?

Not the AI Overviews feature specifically, the answer box embedded directly in ordinary search results, as of this research. Google Canada's own blog states French is "not currently supported" for AI Overviews, with no confirmed date. A related but separate feature, AI Mode, did gain French support globally, including Canada, from October 6, 2025, but it requires a searcher to actively switch to that tab rather than being the default experience.

If I already have a French version of my site, do I still have a GEO problem?

Possibly, yes. Having French pages is necessary but not sufficient. If that French content is a direct translation of English pages rather than content built and reviewed for Quebec French vocabulary and search behaviour, it's likely underperforming both in organic search and in AI-engine citation testing, even though the pages technically exist.

Which AI engine should French-Canadian content prioritise?

Based on NETendances 2025 data on Quebec's generative AI users, ChatGPT carries the largest share at 84%, followed by Microsoft Copilot at 29% and the standalone Gemini app at 22%. A French-Canadian GEO strategy should weight ChatGPT first while still monitoring Copilot and Gemini, and now AI Mode, since Google's default AI Overview surface still isn't part of the French-language mix.

Is this gap likely to close soon?

There's reason for cautious optimism, though no confirmed date. Google has been rolling out French support feature by feature, AI Mode in October 2025, Search Live in March 2026, which suggests AI Overviews specifically could follow a similar pattern. The practical approach is still to build a French-Canadian GEO strategy assuming the current gap persists, while setting a recurring check to re-test AI Overview behaviour for French queries so the strategy can adjust quickly once it changes.

Does Bill 96 apply even if we don't have AI-search goals for Quebec?

Yes. Bill 96's French-prominence requirement applies to any commercial website reaching Quebec residents, independent of whether the business is pursuing AI-search visibility there. GEO strategy and legal compliance point in the same direction here, but they're separate obligations, and the legal one applies regardless of marketing priorities.

---