Who’s blocking AI in Canada? I checked 388 websites
Seven in ten Canadian news sites tell AI crawlers to stay out. Not one bank does. Here’s who’s blocking what, outlet by outlet, with the full dataset to download.
Contents
The bit to remember
- 70% of the Canadian news sites I checked (65 of 93) tell at least one AI crawler to stay out. Across every other sector combined, it’s 5%.
- Not one of the 29 banks, insurers and fintechs in the study blocks an AI crawler.
- French-language newsrooms block the most: 86% block at least one AI crawler, and 82% block OpenAI’s GPTBot, against 39% of English-language outlets.
- Nine sites, including CBC, Radio-Canada and CTV News, block old Anthropic crawler names that aren’t in Anthropic’s current documentation, while leaving its current crawler, ClaudeBot, without a rule.
- 16 sites block Cohere, the Toronto AI company Ottawa committed $240 million to, including CBC, The Globe and Mail, The Canadian Press and the Toronto Star.
Summarize with AI
Ask an AI assistant about mortgage rates in Canada and it’s free to read every Big Six bank’s website, top to bottom. Ask it about this morning’s news and seven in ten Canadian news sites have put up a sign asking it to leave.
In October 2026, I pulled the robots.txt file from 388 of the best-known websites in Canada: national, regional and French-language news, banks, telecoms, retailers, every province and territory, the biggest cities, universities and tech companies. Robots.txt is the small public text file where a site tells crawlers what they may read. Here’s who’s letting AI in, who isn’t, and what that means if you’d like AI to talk about your business.
How many Canadian websites block AI crawlers?
21% of the Canadian websites in this study (78 of the 372 I could read) block at least one AI crawler from the whole site. That average hides a split: 70% of news outlets block AI, while only 5% of everyone else does.
| Sector | Sites read | Block any AI crawler | Block OpenAI’s GPTBot | Block ChatGPT search |
|---|---|---|---|---|
| News and media | 93 | 65 (70%) | 46 (49%) | 28 (30%) |
| Tech and AI companies | 35 | 4 (11%) | 1 (3%) | 0 |
| Travel and transport | 16 | 2 (13%) | 1 (6%) | 0 |
| Retail | 40 | 2 (5%) | 1 (3%) | 1 (3%) |
| Education | 39 | 2 (5%) | 1 (3%) | 0 |
| Government | 63 | 3 (5%) | 1 (2%) | 0 |
| Banking and finance | 29 | 0 | 0 | 0 |
| Energy and industry | 24 | 0 | 0 | 0 |
| Telecom | 15 | 0 | 0 | 0 |
| Food and restaurants | 11 | 0 | 0 | 0 |
| Health | 7 | 0 | 0 | 0 |
| All sites | 372 | 78 (21%) | 51 (14%) | 30 (8%) |
A BuzzStream analysis of the top UK and US news sites published in January 2026 found 79% of them block at least one AI training crawler, and Zyte’s State of Web Access 2026 puts AI blocking across Canadian sites in general at 14%. Neither one names Canadian outlets, splits Canada’s news from its banks, or separates French from English, so that’s what I wanted to find out.
French-language newsrooms are the strictest in the country
French-language news is the most locked-down group in the data, with 19 of the 22 French-language news sites I checked blocking at least one AI crawler. The gap with English-language outlets shows up on every major crawler.
| Crawler blocked | French-language news (22) | English-language news (71) |
|---|---|---|
| Any AI crawler | 86% | 65% |
| GPTBot (OpenAI training) | 82% | 39% |
| ClaudeBot (Anthropic training) | 73% | 28% |
| Google-Extended (Gemini training) | 59% | 27% |
| OAI-SearchBot (ChatGPT search) | 59% | 21% |
La Presse’s robots.txt opens with a bilingual notice saying its content may not be used “for large language models (LLMs), machine learning and/or artificial intelligence-related” purposes, and sends anyone who wants a licence to Copibec, Quebec’s copyright collective. It then blocks more than 50 crawlers, Apple’s regular search crawler included. Le Devoir, Le Soleil, Le Droit, Le Nouvelliste, La Tribune, L’actualité, Les Affaires and Protégez-Vous are full blockers too. The only French-language outlets in the study with no AI blocks are Urbania, 98.5 Montréal and ONFR.
Canada’s media owners each picked a different strategy
Sister sites in a media chain usually share one robots.txt template, so the AI decision gets made once at head office and copied across every paper.
| Owner | Sites in the study | What their robots.txt does |
|---|---|---|
| Torstar, The Globe and Mail, The Canadian Press | Toronto Star, Hamilton Spectator, Waterloo Region Record, The Globe and Mail, The Canadian Press | Blocks almost everything. The Star’s file names more than 200 bots, and The Globe’s names more than 180, including the PR monitoring tools Muck Rack, Cision and Meltwater. |
| Quebecor | Le Journal de Montréal, Le Journal de Québec, TVA Nouvelles, QUB | Blocks OpenAI’s and Anthropic’s training crawlers, plus Claude’s search crawler, but leaves ChatGPT search open, so its pages can still appear in ChatGPT answers without feeding OpenAI’s training. |
| Postmedia | National Post, Financial Post, the Sun chain, Montreal Gazette, Vancouver Sun, Calgary Herald and more | 19 sites point to the same Postmedia content licence file. Six of them (Montreal Gazette, Regina Leader-Post, Saskatoon StarPhoenix, London Free Press, Windsor Star, Kingston Whig-Standard) also block AI crawlers outright. The National Post, Vancouver Sun, Calgary Herald and the others mostly don’t. |
| Bell Media and CBC/Radio-Canada | CTV News, CP24, BNN Bloomberg, TSN, Noovo Info, CBC News, Radio-Canada | Near-identical lists. They block OpenAI (ChatGPT search included), Perplexity and Common Crawl, but their Anthropic rules use older names Anthropic no longer lists. |
| Rogers and Corus | CityNews, Sportsnet, Global News | No AI blocks at all. |
The Postmedia licence file uses Really Simple Licensing (RSL), an open standard released as version 1.0 on December 10, 2025 that lets publishers spell out in machine-readable terms how AI systems may use their content. Postmedia is putting a price tag on the door rather than a lock. Half a dozen of its papers have a lock as well, which suggests the chain hasn’t fully settled on one policy.
Quebecor isn’t alone in splitting training from search: 19 news sites block GPTBot but allow OAI-SearchBot. OpenAI’s crawler documentation treats those as separate decisions: blocking GPTBot “indicates a site’s content should not be used in training generative AI foundation models”, while “sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers.” The 28 news sites that block OAI-SearchBot, CBC and CTV News among them, are opting out of ChatGPT search answers altogether.
Nine sites block Anthropic crawler names that are no longer listed
AI companies rename and add crawlers often, and plenty of robots.txt files haven’t kept up. The clearest case involves Anthropic, the company behind Claude.
Anthropic’s current crawler page lists three bots: ClaudeBot for training, Claude-SearchBot for search indexing and Claude-User for fetching pages when someone asks Claude a question. The names anthropic-ai and Claude-Web don’t appear on it. Nine sites block only those two older names and have no rule for ClaudeBot:
- CBC News and Radio-Canada
- CTV News, CP24, BNN Bloomberg, TSN and Noovo Info (Bell Media)
- blogTO
- George Brown College
15 sites block OpenAI’s GPTBot without blocking ClaudeBot. That could be deliberate, for example as part of a licensing arrangement, or it could be a template nobody has opened since 2023. The file can’t tell you which, so I’d treat these as questions for the publishers rather than mistakes.
Banks, governments and retailers are leaving the door wide open
- Banks: None of the 29 banks, insurers and fintechs blocks an AI crawler. That includes RBC, TD, Scotiabank, BMO, CIBC, National Bank and Desjardins.
- Government: 60 of the 63 federal, provincial and municipal sites block nothing, canada.ca included. The exception is the Government of Manitoba, which blocks 11 of the AI crawlers I tracked, including GPTBot, ClaudeBot, Google-Extended, PerplexityBot and Cohere’s crawler.
- Retail: 38 of 40 retailers are open. The one full blocker is Amazon.ca, which blocks 19 of the 22 AI crawlers I tracked (its own Amazonbot is allowed), though Amazon is an American company with a Canadian storefront.
- The welcome mat: Seven sites point AI systems to an llms.txt file, a newer convention for handing AI a clean summary of a site: TD, Tangerine, belairdirect, Canadian Tire, Sport Chek, Mark’s and Jobber.
When a large share of the country’s journalism asks AI crawlers to stay out while brands, banks and governments let them in, what AI tools can freshly read about Canada tilts toward the people selling something. Robots.txt doesn’t remove anything a model has already been trained on, and licensing deals can bring blocked content back in, so this is a tilt, not a blackout.
Cohere is blocked by the papers that cover it
Cohere is the Toronto-based AI company the federal government committed $240 million to in December 2024 to expand its computing capacity. 16 of the sites in this study block its crawler, including CBC News and Radio-Canada, The Globe and Mail, The Canadian Press, the Toronto Star, the Hamilton Spectator, The Hill Times, MobileSyrup and the Government of Manitoba.
Canadian sites aren’t giving the home team a pass. Four of them (MobileSyrup, blogTO, Daily Hive and Métro Montréal) block Cohere while letting OpenAI’s GPTBot in.
Most businesses should do the opposite of the newsrooms
A publisher sells its words, so letting AI read them for free can cost it money. A plumber, a law firm or a supplement brand mostly wants to be found, and an AI assistant can only recommend what it’s allowed to read.
Sites can also block AI by accident. A template from 2023, an overcautious plugin or a network-level bot setting (Cloudflare offers one) can shut out the crawlers that decide whether you show up in ChatGPT, Claude or Perplexity answers. You can check yours in about two minutes:
- Open yourdomain.com/robots.txt in a browser.
- Search the page for GPTBot, OAI-SearchBot, ClaudeBot, Claude-SearchBot, PerplexityBot and Google-Extended.
- If any of them sit above a
Disallow: /line, that crawler is blocked from your whole site. - Look at the
User-agent: *group too. ADisallow: /there blocks every crawler that isn’t named elsewhere. - If you use Cloudflare or another firewall, check its bot settings as well, since those can block AI crawlers that robots.txt allows.
If you want to stay out of AI training but stay visible in AI search, which is the Quebecor approach, the robots.txt rules look like this:
# Keep content out of AI model training
User-agent: GPTBot
User-agent: ClaudeBot
User-agent: Google-Extended
User-agent: Applebot-Extended
User-agent: CCBot
Disallow: /
# Stay visible in AI search and answers
User-agent: OAI-SearchBot
User-agent: Claude-SearchBot
User-agent: PerplexityBot
User-agent: ChatGPT-User
User-agent: Claude-User
Allow: /Google says Google-Extended only controls whether content is used to train Gemini, and that it “does not impact a site’s inclusion in Google Search nor is it used as a ranking signal”. Blocking it doesn’t cost you Google rankings.
Being readable only gets you in the door, since AI tools still have to trust your pages enough to cite them. That takes work: content that answers the questions your buyers actually ask, and mentions of your brand on sites AI already relies on. If you’d like to know where you stand, my AI visibility audit checks both. And if you’ve just discovered you have no idea what an llms.txt file is, llms.txt explained covers it.
How I ran the study
I fetched each site’s robots.txt live on October 17, 2026 and parsed it against the robots.txt standard (RFC 9309). For each of 22 AI crawler names, the parser finds the group of rules that names that crawler, or falls back to the catch-all * group the way the crawler would. A crawler counts as blocked only if it’s barred from the whole site: the homepage, a deep article URL, /en/ and /fr/. Blocks on single folders don’t count.
- The sample: 388 sites across 11 sectors, picked to cover the largest and best-known Canadian organizations in each. It’s a structured sample, not a random one.
- Excluded sites: 16 sites couldn’t be read because the fetch failed or returned a web page instead of a robots.txt file, including Air Canada, Sun Life, Tim Hortons and the Senate. They’re left out of every percentage.
- No file: sites with no robots.txt file at all count as open, since that’s how crawlers treat them.
- “Full AI blocker”: blocks at least 6 of the 7 major training crawlers (GPTBot, ClaudeBot, Google-Extended, Applebot-Extended, Meta-ExternalAgent, CCBot and Bytespider).
Robots.txt is a voluntary request. It shows what a site asks crawlers to do, not what crawlers actually do, and it can’t show blocking done at the network level. These files change often, so treat this as an October 2026 snapshot.
The full dataset, with every site and every crawler, is free to use with credit to Chateau Mojo. Journalists who want the raw robots.txt files or a cut of the data for a story can get in touch.
FAQ
Which Canadian news sites block AI crawlers?
Most Canadian news outlets block them. 65 of the 93 Canadian news sites in this study block at least one AI crawler. The strictest include La Presse, Le Devoir, The Globe and Mail, the Toronto Star, The Canadian Press and the Montreal Gazette, which each block nearly every major AI training crawler.
Does blocking GPTBot stop ChatGPT from citing a site?
No. GPTBot only controls whether OpenAI uses a site’s content for model training. ChatGPT search uses a separate crawler, OAI-SearchBot, and OpenAI says sites that block OAI-SearchBot “will not be shown in ChatGPT search answers.”
What’s the difference between GPTBot and OAI-SearchBot?
GPTBot collects content for training OpenAI’s models, while OAI-SearchBot indexes pages so they can appear in ChatGPT search results. Blocking one doesn’t block the other, which is why 19 Canadian news sites block GPTBot and still allow OAI-SearchBot.
Is robots.txt legally binding in Canada?
No. Robots.txt is a voluntary standard that well-behaved crawlers follow. Some publishers pair it with terms of use or a licence notice, like La Presse’s bilingual statement, but the file itself doesn’t enforce anything.
Should my business block AI crawlers?
Probably not, unless your content is the product you sell. Most businesses want AI assistants to recommend them, and an assistant can’t recommend what it isn’t allowed to read. If you’re uneasy about AI training, you can block training crawlers and still allow AI search crawlers.
How do I check whether my website blocks AI?
Open yourdomain.com/robots.txt and look for AI crawler names such as GPTBot, ClaudeBot and PerplexityBot above a Disallow: / line. Check your firewall or CDN bot settings too, since services like Cloudflare can block AI crawlers that robots.txt allows.
Why do newspapers block AI but banks don’t?
Newspapers sell their content, so AI companies reading it for free competes with subscriptions and licensing deals. Banks, retailers and governments want people to find their information, so being read and quoted by AI tools helps them.
