Is Cloudflare Blocking AI From Your Website? We Found Out the Hard Way

Question 01

When did you first notice traffic had dropped? Or did you not know until you turned the filters off?

That's the uncomfortable part: there was no drop to notice. The blocking was on from the day the site moved to Cloudflare in November 2025, so there was never a before and after on my charts. Google was never blocked, and our Search Console graph climbed all year. Everything looked healthy.

What actually tipped us off was the opposite of a drop. On June 24, 2026 we noticed an anomaly: a surge of traffic from Singapore that made no sense for a Brooklyn law firm. We went digging to find out what it was, and pulling that thread through the Cloudflare data is how we stumbled onto the block. We went in looking for bogus traffic and found out the real visitors, the AI crawlers, were the ones being turned away.

You can't see traffic that never arrives. The only trace was Cloudflare's dashboard counting blocked crawler visits as threats, which reads like the product doing its job. We had no idea for seven months.

Question 02

What made you go into Cloudflare and change the setting? Did someone tell you, or did you figure it out?

Nobody told me. The Singapore anomaly sent us digging, and on July 3rd, 2026, as part of that investigation, I was running a search-visibility audit with Claude, the AI assistant I use for the firm's website. We tested whether the major crawlers could actually read the site. Googlebot came back fine. GPTBot, PerplexityBot, and Common Crawl all came back with a 403 error from Cloudflare.

Cloudflare's own panel confirmed it: in the previous 24 hours, GPTBot had 0 requests allowed and 515 blocked. PerplexityBot, 0 allowed and 288 blocked. It was a default setting, not a decision anyone at my firm ever made. I flipped the toggles that afternoon, July 3rd, 2026.

Question 03

What happened to traffic after you turned off the AI crawler filters?

The next day told the whole story. GPTBot went from 0 requests to 14,292. PerplexityBot went from 0 to 7,483 and became the single biggest crawler on the site. OpenAI's search index crawler showed up for the first time. Total traffic jumped 29 percent in one day.

It held: page loads went from about 25,500 a day to about 37,200, up 46 percent, and blocked threats fell 98 percent. The number I watch now is ChatGPT-User, the crawler that only fires when a real person asks ChatGPT a question and it reads your site live to answer them. It hits us 400 to 1,600 times a day. Before July 3rd, 2026, OpenAI could not index us at all.

Question 04

Can you tie the traffic increase to actual revenue? New clients, consultations, cases?

Yes. New clients are telling us to our faces that they found us through ChatGPT or another AI tool. We ask every new client how they heard of us, and that answer keeps coming up more often. And the money moved with it: measured per day, client receipts rose 11 percent from May to June 2026, and the first twenty days of July are up another 28 percent on top of that, 42 percent above the May baseline. July is pacing to be the strongest month in any window we can measure. We keep the dollars private, but the direction is not subtle.

The traffic from those same weeks tells you where the attention came from: our best organic search week on record (747 Google clicks, up 13 percent week over week), ChatGPT reading the site as many as 1,600 times a day, and 126 visitors clicking through from AI tools in four weeks, versus 36 in the five months before.

Could I trace every dollar to a single source? No, and I'm a lawyer, so I won't pretend the ledger has one author. The ads, the content, referrals, the season, they all contributed. But the phone tells the story the dashboards can't, because impressions don't just turn into clicks, they turn into calls, and the people calling are naming the AI tools.

My read as the owner is that it's everything compounding. Our ads catch the people who search. AI visibility catches the people who ask. For seven months that second door was closed, and now ChatGPT consults our site hundreds of times a day to answer people asking about exactly the problems we handle. People ask, the AI answers, and when that answer comes from our site, our phone rings.

Question 05

How much revenue were you leaving on the table? Can you estimate an annual number?

Let me say the part I can prove first: revenue went up. Client receipts per day rose 11 percent from May to June 2026, and the first twenty days of July are up another 28 percent on top of that, 42 percent above the May baseline. July is pacing to be the strongest month in any window we can measure. What I won't do is invent an annual number for the seven months we were blocked; that exposure was real, and it went unmeasured, which is exactly the problem.

The math is simple once you start tracking it: how many people came from AI tools, how many of them call, how many become clients, and what a case is worth. Our real July 2026 numbers: 126 visitors from AI tools in four weeks, ChatGPT reading the site hundreds of times a day, and cases that routinely run five figures. It does not take many of those conversations turning into clients to matter. That door was locked shut for seven months.

Question 06

Did Cloudflare ever notify you these filters were on?

No. No email, no alert, nothing in the dashboard flagged it. The blocking came switched on when the zone was created, and it also quietly rewrote our robots.txt with an AI opt-out signal we never chose.

We found it because we went looking. If we hadn't run that audit through Claude on July 3rd, 2026, it would still be on today, quietly turning away every AI tool our future clients ask.

Question 07

What would you tell other small business owners who are on Cloudflare right now?

Owner to owner: don't just invest in regular SEO. The way people find help is changing under our feet. They don't only search anymore, they ask. They ask ChatGPT who can fight their violation, they ask Perplexity what a closing costs, and whoever the tools can read becomes the answer. That's answer-engine optimization, AEO, and if you're only optimizing for Google you're competing for a shrinking share of the attention.

Don't underestimate what this is worth. We're a small firm. No agency, no marketing department. We wrote useful content, we opened the door to the AI crawlers, and we saw a marked increase in traffic: page views up 46 percent, ChatGPT reading us hundreds of times a day, and new clients telling us on the phone that an AI tool sent them. None of that required a budget. It required being readable.

The mechanical part takes five minutes: assume you're blocked until you prove otherwise, test it by asking ChatGPT to summarize a page of your site, and know the fix is two toggles in Cloudflare's AI Crawl Control panel. The step-by-step is in the checklist below. But the mindset is the real advice: the next customer is already asking. Make sure the thing they're asking can read you.

Does Cloudflare block AI crawlers by default?

For zones created since mid-2025, yes. New Cloudflare sites ship with AI-bot blocking switched on, and a separate managed robots.txt feature can inject an AI opt-out signal into your robots.txt without your input. Our zone was created in November 2025 and both were active from day one. We were never notified either was on.

Will I see the block in my analytics?

No, and that is the trap. Visitors and crawlers that never arrive do not show up as a drop in any chart. Our Google rankings and traffic kept climbing the entire time the block was on. The only trace was Cloudflare counting the blocked crawler visits as threats, which looks like the security product doing its job.

How do I check if my site is blocked?

The thirty-second test: ask ChatGPT or Perplexity to summarize a specific page of your website. If it cannot read the page, check Cloudflare under AI Crawl Control. The fix is two toggles: allow the crawlers you want in the Security tab, and review what Managed robots.txt is publishing in your name under Signals.

Should every business allow AI crawlers?

Not necessarily. If your business sells the content itself, blocking AI training can be a rational choice. For a service business that wins clients by being found, the calculus is different: when someone asks ChatGPT who handles their problem, you can only be the answer if the tools can read your site. Either way it should be your decision, made on purpose, not a default you never saw.

What did seven months of blocking cost?

We cannot put an honest dollar figure on it, and we will not invent one. What we can measure: within two weeks of unblocking, ChatGPT alone was reading our site 400 to 1,600 times a day to answer real people's questions, and human visits referred by AI tools were arriving at roughly nineteen times the old daily rate. Every one of those conversations was structurally impossible the day before we flipped the toggles.