I pulled Figma's robots.txt this morning expecting a rumor and found a manifesto. There it is, in plain text: Google-Extended, Disallow: /. And it's got company: GPTBot blocked, ClaudeBot blocked, CCBot, cohere-ai, and FacebookBot all shown the door, every one of them barred from the entire site. Meanwhile regular Googlebot strolls in like always. Figma blocking Google AI is real, it's deliberate, and it's one of the most instructive robots.txt files in SaaS right now, mostly because of what it teaches everyone who shouldn't copy it.
What Figma Actually Blocked (And Didn't)
Let's decode the file, because the details carry the lesson.
Backing up a step: Google-Extended is Google's AI training token. Blocking it tells Google: don't train Gemini-family models on this content. Here's the thing, though: it does no more than that. AI Overviews and AI Mode run on Google's core Search systems and Googlebot's crawl, and Googlebot is welcome at Figma. So Figma keeps its rankings and Search presence, and can keep surfacing in AI Overviews, all while opting its content out of Google's model training.
The other blocks go further. GPTBot and ClaudeBot are how OpenAI and Anthropic crawl the web for their engines, so barring them means those systems can't read figma.com at all. CCBot cuts off Common Crawl, the dataset half the industry trains on. That's a real lockout, not a symbolic one.
Why Figma Gets Away With It
Real talk: Figma could block every crawler on earth and your designer would still hear about Figma this week.
The models already know what Figma is. They learned it from years of tutorials, YouTube walkthroughs, community forums, comparison posts, and every design-Twitter argument ever archived. Buyers ask for Figma by name, which means AI engines answer questions about Figma using an ocean of third-party content that Figma doesn't control and can't block. Withholding figma.com from training barely dents any model's understanding of Figma; it mostly denies the labs free training data, which for a company sitting on a mountain of proprietary design content is a defensible trade.
That's brand gravity, and it's the entire trick. The blockade works because the brand no longer depends on its own website to exist in machine minds.
We'll show you exactly where your brand stands in AI search.
No commitment. $0 due today, cancel anytime. See how ChatGPT, Perplexity, Claude, Gemini, Grok, and Google AI Mode talk about your brand today.
Why You Probably Shouldn't Copy It
Now the part I actually care about, because I keep meeting founders who read a headline like this and reach for the deny list.
If your brand lacks household gravity, blocking search-and-cite crawlers protects nothing and silences the one voice you control. The engines answering your buyers' questions will still answer them, using whatever third parties say about you: the outdated review, the competitor's comparison page, the directory listing nobody claimed, the Reddit thread from two pricing models ago. You've stepped out of the conversation and left every voice you don't control still talking. Blocking the crawlers that feed citations when nobody's asking for you by name is how a brand becomes a rumor.
And spoiler alert: the trade Figma made isn't even available to you. They gave up training exposure they didn't need. You'd be giving up discovery you very much do.
The Numbers Behind the Third-Party Truth
The research on citation sources backs this up with an uncomfortable ratio: industry studies have found brands are several times more likely to be cited in AI answers through third-party sources than through their own domains. Your site was never the main character in how AI describes you; it's one witness among many, and often a minor one next to Wikipedia, YouTube, review platforms, and community threads.
The publishing industry is running the control experiment in public. Legacy paywalled press, which locks crawlers out as policy, is conspicuously missing from ChatGPT's most-cited source lists, while nearly 300 French newspapers spent the summer fighting Google over AI Overviews through regulators instead of robots.txt. Walls keep the crawlers out. They've never yet kept a brand out of the answers, only out of the room where the answers get written.
Check Your Own File Before You Have Opinions
One more Casual-Friday exercise, and I mean this literally: go read your own robots.txt right now. Load yoursite.com/robots.txt and search it for GPTBot, ClaudeBot, PerplexityBot, CCBot, and Google-Extended.
I keep finding SaaS sites that block AI crawlers without anyone on the team remembering a decision. Copied robots templates, CDN bot-protection defaults, a security vendor's recommended file from 2023: accidental policy accumulates, and it produces exactly the silence described above, minus the strategy. If you find a block you can't explain, that's not a Figma move. That's a leak, and it's costing you answers today.
The Middle Path Most SaaS Should Run
So what's the grown-up version of this decision? Split crawlers by function; it takes twenty minutes.
- Keep search-and-citation bots open. GPTBot, ClaudeBot, PerplexityBot, Googlebot: these feed the answers your buyers read. For a growing brand they're distribution, full stop.
- Decide separately on training-only crawlers. CCBot and Google-Extended are the genuine training levers. Blocking them is a philosophical choice with modest visibility cost, and it's the only part of Figma's playbook that transfers safely.
- Write it down and re-check quarterly. Crawler tokens multiply and the allow/block list changes faster than anyone's robots.txt habits. An unowned robots.txt accumulates policy by accident.
- Monitor the answers regardless. Blocked or open, AI engines keep describing you, because their raw material was never only your site. Watching what they say weekly is the instrument that tells you whether your crawler policy is working or quietly starving your presence. RankControl's agents run that watch across the major engines continuously, which is exactly the feedback loop a decision like this needs.
Know exactly what AI says about your competitors.
RankControl's Recon Agent monitors competitor citations across ChatGPT, Perplexity, Claude, Gemini, Grok, and Google AI Mode. See where they show up and you don't.
The Actual Lesson
Funny enough, the takeaway from Figma blocking Google AI has almost nothing to do with robots.txt syntax. Figma spent a decade becoming so present in the web's conversation that its own site became optional to its reputation. That's the position that makes blocking affordable, and you earn it through the unglamorous work: being taught, discussed, compared, and recommended everywhere, until the machines learned you from everyone else.
Get famous first. Then you can afford to be private. Do it in the other order and the machines will describe you anyway, using material you'd never have chosen.
26 content formats. Published on your domain. Matched to your brand.
Guides, comparisons, listicles, case studies, and more. RankControl generates content that gets cited by ChatGPT, Perplexity, Claude, Gemini, Grok, and Google AI Mode.
