Two things in Cloudflare's second Content Independence Day report land on a marketing leader's desk. The traffic numbers you report are now half machine, and a September setting changes how the biggest AI companies reach your pages.
Strip out the infrastructure talk and Cloudflare's new report says two things a marketing leader can act on this week. More than half the traffic hitting websites is now automated software, not human beings. And the visitors who still count, the people who arrive and might buy something, mostly come from Google, not from the AI chatbots everyone spent the year worrying about. Cloudflare published this in its second Content Independence Day report on July 1, and it runs a large share of the world's websites, so the numbers come from real traffic even though they are the company's own figures.
Here is what each of those means for you.
The traffic report you send upstairs is counting robots as visitors
Every month your team pulls a number for how many people visited the website. That number is now inflated, because more than half of what shows up is software, not shoppers. Bots do not fill out a form, watch a demo, or place an order. They just add to the count. So when your dashboard says traffic is flat but sales inquiries are down, the two are not in conflict. The human part of your traffic is shrinking while the robot part grows and hides the drop.
The fix is a checkup, not a project. Ask your web analyst two questions. First, is bot filtering turned on in our analytics, so the reports we act on count people and not machines. Second, where are our human visitors coming from this year compared to last. That second answer is changing fast, because when someone asks an AI a question and gets the answer without clicking through to your site, you never see that visit at all. Your influence is growing in places that leave no trace in your traffic report.
Not all bots are the same, and now you can tell them apart
The old choice was all or nothing. You either let the artificial intelligence bots in or you blocked them, with one switch for everything. Cloudflare has now split them into three kinds, because they do very different things to your business.
Search bots look at your pages so an engine can point people to you later. Those send you visitors, and you want them. Agent bots fetch a page in the moment because a real person asked an assistant to go get something, so a human is waiting at the other end. Training bots copy your writing to teach an artificial intelligence model, and they give you nothing back in traffic. Three jobs, three different answers, and now three separate settings instead of one blunt toggle.
The September setting has a wrinkle that touches Google
On September 15, Cloudflare turns on a new default for websites that newly join its network. On any page that shows ads, it will automatically block the training bots and the agent bots and keep the search bots. The thinking is that an ad means the page was built for a person to see, so keep the machines off it. Websites already on Cloudflare get advance notice and can opt out before that date. New sites get the setting unless someone changes it.
Here is the wrinkle. Most artificial intelligence companies use one bot for search and a separate bot for training, so you can wave in the helpful one and turn away the other. Google uses a single bot that does both at once. Because that one Google bot is part search and part everything else, a setting that blocks the other kinds on your ad pages can reach Google too. And Google is still the source of nearly all the visitors websites get. If your business depends on people finding you, that ad-page setting is a decision to make on purpose, not a default to leave sitting there.
Google sends most of your visitors and uses one bot to do everything, so any rule about bots is quietly also a rule about Google.
Lock the paid rooms, keep the front door open
If some of your content is behind a paywall or a login, gate that. A subscriber database or a paid report is yours to protect, and charging for it is fair. Your blog, your news, your product pages, and your help articles are a different matter. Those are how customers and the tools they use find you in the first place, so keep them open to the well-behaved bots, including the ones that build artificial intelligence answers. A page a machine cannot read is a page that disappears from the search results and assistants people now use to start looking. Block the rooms you sell tickets to, and leave the lobby open.
A number worth watching if you make content
For publishers and any company whose product is its writing, the report carries a useful new figure. Cloudflare's dashboard shows a simple ratio: how many times an artificial intelligence company's bot takes from your site versus how many readers it sends back. Those run from about 118 visits taken for every one sent, up to nearly 50,000 taken for every one sent (Cloudflare; 2026). A lopsided number tells you which companies lean hard on your work and return almost nothing, which is exactly the evidence you would put on the table if you ever ask them to pay for access. Cloudflare is also opening a way to charge those companies, so the block-or-allow choice is turning into a price.
Before September 15, get one honest answer for your own site: how much of our traffic is really people, and how much of that comes from Google. If your current reports cannot separate the humans from the bots or show where the humans came from, that is the first thing to fix, because you cannot make a smart choice about blocking, allowing, or charging any bot until you know which of your visitors were ever real.
Lee, Jin-Hee, and Bryan Becker. "Your Site, Your Rules: New AI Traffic Options for All Customers." The Cloudflare Blog, 1 July 2026, blog.cloudflare.com.
Lee, Jin-Hee, and Oliver Payne. "Unmasking the Crawls with Attribution Business Insights." The Cloudflare Blog, 1 July 2026, blog.cloudflare.com.
