Websites are closing their doors to AI agents
theverge.com
theverge.com
It really sucks for small businesses. Their websites get absolutely hammered and they can’t afford or don’t know how to stop it. Cloudflare ends up winning as they’re the easiest solution.
What really sucks is e.g. Gemini that pulls images without being able to go to the source (unless it changed recently.)
Edit: I didn't read "logs"
Any answer to that along the lines of "do something more complex" or "optimize your site for load" to be able to deal with the rampant bad behavior of multi-billion dollar companies is unpalatable to most.
This is one of those things that might be best solved through policy and law. There probably will be some changes to policy and law to help, eventually. That stuff usually takes a while, and that's always been a problem for those suffering in the interim.
We started with caching but this created a new kind of problems of its own (veterans know what circus of man eating from outside out clowns this create) but client was mad manual refresh is causing "freezes" in the system.
So we ban LLM and problem solved until client starts asking why ChatGPT, Claude, whatever he prefer can't access data and release him/her from pain of using brain
phpBB will assign you a randomized session if if you visit it anonymously (without logging in), and will serve the page to you with that sid param added to all links. Using that link will load a page without the sid param appended to links, because the system assumes you have a session I'd saved in a cookie since it saw you have a sid, even I'd you don't have/didn't send a cookie with it. Following any of those links starts the chain over if you aren't sending a cookie along with a sid, which the AI scrapers never do so they can't easily be blocked by that, and a new unique sid is generated, and all the links have the new sid appended.
Every page with a unique set of params and values needs to be investigated, because it can be an entirely different page, so AI scrapers will scrap forever, always finding more and more URLs to harvest, even though they're all duplicates of what they already have. It's easy to have multiple scrapers stuck going through a very small bulletin board continuously.
Maybe they've gotten smart enough to deal with this situation in the last few months, or maybe they put in a specific workaround. I ended up just throwing the site behind cloudflare because spending time trying to outsmart AI agents in a battle of escalation doesn't seem like an effective use of my time.
Requires login
Bunny.net is another one that I find similar to Cloudflare from pricing and functionality perspective
I've been wondering to myself "how can I get people to go through a login process without going through a login process?" for a few months, and I've come up with a simple solution which I'll be implementing in the next few weeks.[1] Like the other technical measures it doesn't identify bots but there will be a gate with backpressure that at a point in time a given IP address has some probability (adjusted in realtime based on $things) of passing through, and through my mysterious process is then "logged in". Then that address is pretty much given a free pass as long as its behavior remains "human". Real humans may be denied access initially and have to try again, and their luck will be affected by how aggressively the bots are fighting it out at that point in time. This isn't much different than the extant reality, but it is in terms of what happens after an initial successful page load. It's an experiment, I'm hoping it's tunable so that the experience is more or less indistinguishable from the status quo; we will see.
I'd love to find a group of likeminded individuals where we can discuss some of this TLP:YELLOW or Chatham House Rules. If you are savvy enough to hunt me down and can demonstrate that you're a real human, small operator, and that you can keep things closely held I'll talk with you, if you want someone to run your mitigations that would cost money.
[0] Had an episode last week where somebody scanned approximately 1000 URLs (maybe 60% of the discoverable URLs) over six hours. 1200 addresses fetched exactly one URL, another 400 or so fetched two or three. Every one of those URLs was fetched exactly two times. Real showshoe stuff, nicely done.
[1] Firewall rules and fail2ban. I give fail2ban special powers in terms of the logs it follows and the actions it can invoke. Conntrack without conntrack (of course the firewall has conntrack, it's just not the best option for this use case).
What prevents other hosting sites from providing similar tooling?
As a user (bot or flesh) you have to jump through these hoops, and to the website operators they get fewer complaints when users (and bots) are able to contort themselves in predictable ways. Versus you'd perceive as a website operator, if your blocking worked at all, that you'd risk seeing a precipitous dropoff of bots if you did things in bespoke ways, as both bots (true positives, among dd and cf survivors) and people (false positives) get jammed.
The net effect in this model would be that such things as cloudflare and datadome take money from websites in order to block "dumb" bots and enfranchise "smart" bots to the point of creating a market for bot services. People that want to bot have to use AI or go to bot service providers.
For website operators that's a worthwhile exchange for dollars, in that dumb bots can cause a lot of harm in wasted compute. Bots are typically dead end users who won't buy ever anyway, or merely inefficient way of doing business if they scrape excessively. And many do scrape excessively.
"reform school for bots" is a good mental model. For this to work, the "reforms" have to be consistent so that bots can conform to them.
It's a peculiar equilibrium and I am not sure that it's all that stable as an equilibrium. Would be a good subject for a formal treatment, game theory and economics.
It works similar to cloudflare WAF but is an open source software. It can be on your own infrastructures even if its a $5 VPS.
The idea is same let the bot or AI agent do more work than it takes to serve that request. Also it needs to follow rules, similar to cloudflare enterprise web application firewalls. It was OWASP CRS and J4A TLS support.
Considering sibuna it's anubis backwards.
Look like rust rewrite to make more performant?
And the problem often isn't some advanced attack. It's enough to follow links with session params, generate endless urls or ignore robots.txt to turn a regular crawler into a load generator. A site owner needs crawler protection from specific behavior and not another list of known bot signatures
Another more blunt way to say it is that people can instruct their agents to look out for their best interests (and while there are cracks they continue to get better at this), and companies hate that.
People will work around it similar to how we've worked around with adblockers for years, and it will just become a game of attrition as these companies work to preserve their moats until they find a way to preserve this revenue stream off of people's agents.
I do see that there is another side to this. I saw complaints by an event organizer (think weddings etc) who got some special requests and went above and beyond to accommodate but it turns out the extensive requests likely were AI generated fully optimizing for the customer's interests, and just shamelessly asking for many favors essentially. She used to do these things because humans had a bit of embarrassment and friction around asking too many things and so it worked on a social agreement that if you ask it likely is truly important that you made the effort to ask. Now this cannot be assumed. So things will more to more formal rules, more explicit pricing for the small asks, and generally more adversarial and low trust systems that fully squeeze out the human slack from these systems.
A year ago, I contacted DataDome directly and they assured me that "fighting false-positives is their biggest priority" and their filters were fixed for a while until recently. A few days ago, I contacted them again, and now their stance is different: "our support channel is for customers only and this ticket will be closed." However, their customers have no idea that DataDome messes with their end-users, and reaching a tech support person who actually knows about their website using DataDome is unbelievably difficult.
Datadome reminds me of Cloudflare in their early days, when they were breaking UX for many users with Google reCAPTCHA.
Disclosure: I work at friendly captcha.
Want to argue for a shining new world, but you're on a locked site? I guess we'll have to live with dystopia. Want to sell the greatest widget since the invention of sliced bread? Well, I never found it. Sorry, I bought inferior widgets from your competitor instead. I'm at a physical sales location, but I can't access your spec-sheets? Your competitor the next shelf over doesn't have that problem, I'll go buy theirs.
I'm trying to build a personal social media client to stop the platforms from shoving their feeds down my throat and to consume only the content I care about, so this hits home.
(I just got blocked from LinkedIn for letting an agent set all of the 77 email notification toggles off, which in hindsight was obvious but I wasn't going to do that manually since every toggle has their own separate page, it's a lot of clicks.)
It might not hit input cache at that rate but better than a block or ban. Keep the context window small.
"Companies" are not capable of feeling irony. They absolutely will demand the unfettered right to scrape your stuff while demanding that your scrapers be outlawed at the same time, and if you go "hey, that's hypocritical of you", you won't get any response, not even a half-hearted rebuttal or uncaring shrug. Taking about "hypocrisy" or "philosophical consistency" with regard to corporate behavior is almost a category error, these aren't concepts which even exist in that space.
Walmart is the only other company that can compete at that scale given their very mature ground distribution network and even then Amazon has them beat good on air cargo (Amazon has a virtual cargo airline operated by Air Transport/Atlas Air, both of whom they have some ownership over) and inventory management (they hired some very very smart folks to help them build inventory forecasting systems)
As long as Amazon remains the only company that you can buy almost anything and get it in two or three days almost anywhere on Earth, they won't need to worry about the grim reaper coming for them.
Even if they could, it will be a _long_ time before the average person uses an agent for personal tasks like this
If I have to drive to the store to get it, I wouldn’t have shopped with them! It feels like every time I don’t order with Amazon the stores remind me why I ought to.
Now whether you want to restrict your shop to only being accessed via a mobile app is a totally different proposition.
Agents invalidate a lot of that and so it is an existential threat to many of them. I expect the near future will result in a lot of different results tailored to agents that may actually result in higher priced products for agents browsing versus humans.
https://docs.lightning.engineering/the-lightning-network/l40...
Fortunately, if AI is really superhuman in capabilities it may still win the battle for the customer and keep prices competitive: aggressive price comparison, alternatives discovery, negotiation, group buying, etc.
First they forced their malware bot farms on us, now they want to rent them to us so they can turn the entire web into the dark web.
And eventually rug pull off the major retail sites and just redirect traffic to their own darker sites to scrape commissions, sell direct and keep the change.
I'm sure the teams thinking this up buy their lunches from a ghost kitchen run by illegals that's never had a health inspection,
located in the back of a laundromat with a phony business license and website that only accepts crypto.
Both the consumer and the manufacturer/labor/driver/gig-worker always get the short end of the stick.... once AI-agents becomes indispensable part of life.
Either way, the benefits of price transparency are not imaginary, but they are also not the only factor. Things like insufficient buyers or sellers, government policy, etc are also in play.
All else being equal, a buyer being able to easily see seller A is more expensive than seller B for the same goods will result in lower prices for the buyer. But all else is rarely equal.
It would be way different if it was an open network where drivers are allowed to set prices and reject rides without any penalty, and riders can send multiple offers at a time then take the best one. That would be a real competitive market where everyone is free to choose what they buy and sell.
In that scenario Uber would be one of many apps you can use that index the network, whitelist trusted drivers and handle payments for a small fee. And if you want to make a new app or do the whole process manually there's nothing to stop you. That's what shopping on the open web is like.
> Instead of trying to fight Amazon, the company has partnered with retailers like Walmart, Gap, Sephora, Best Buy, and others, allowing Muse to shop on their sites. At the same time, Shopify has also opened its network of storefronts to AI agents.
they're making deals with the sites so that the agents can work, the sites get compensated. I hope for this "AI can do what it wants if it pays" model to expand further. Cloudflare told us months ago we'd have an option to get paid by AI crawlers. I signed up to join but of course as a nobody I never got a reply.
So the stores aren't being paid to let the agent in. Meta expects them to pay it.
Put another way, because it undermines the revenue that is providing you the service. Working around it is not a sustainable solution.
Not outlandish. The prices are a big motivator and I wonder how much they’d have to charge me to compensate. My agent probably sees the sponsored results before it sees the other results but that is probably considered insufficient compensation.
I think I’ll keep doing what I’m doing but primarily because if the world is forced to charge higher prices to compensate for the lack of advertising I think that isn’t an outcome I particularly am upset by. It passes a Categorical Imperative Test, if you will.
Email was awesome when it first became widespread. I could cold-email random professors at other universities (I was in school at the time) and get meaningful responses back. Then lots of people (and companies) started cold-emailing, and it stopped working.
Social media is broken if there's nobody there, is great with a medium-sized community, and is broken if everyone is there.
Ad-supported websites can offer all kinds of valuable free services. Up until either (1) they become so ad-infested that they're no longer useful, or (2) all the users block ads.
There was a wave of websites providing APIs for programmatic access a while back. You could script together whatever flow you wanted. If no API was offered, you could scrape the HTML. That triggered a wave of alternate interfaces to sites, skipping their ads, skipping their engagement funnels, separating the value from the cost. So everything got locked up behind captchas and accounts and API keys with limited access.
This outcome was always inevitable. Humans don't understand social contracts, or they do and both sides abuse the hell out of it. We feel entitled to have just the loss leaders without ever buying anything else. When running businesses, we feel entitled to keep adding more and more adds and generally enshittifying the experience. People can take it or leave it, after all.
Agents don't change the underlying structure at all. They just magnify and accelerate parts of the system, and will therefore magnify and accelerate the reactive and defensive responses.
Look at spotify and their royalties. Theres already such little value in human based internet that agents will race to the bottom.
The only internet content will be worse than adware. Itll just be agent poison meant to take your money.
Any other viewpoint seems to ignore the already shattered shared internet.
I constantly get pestered by "are you a bot" queries.
They ruined the world wide web here. I did not have that in the 1990s or early 2000s.
Wasn't there some restaurant review site that gets accused of doing pay-to-play hiding of negative reviews and promoting positive ones that calls up businesses and basically extorts them into their services? They have all the incentive in the world now to manipulate the results shown to AI agents. You won't even see the manipulation directly yourself when there's an AI sitting between you and the source acting like an obfuscation layer.
Grim. No point in writing blog posts or publishing anything for free if that's the future.
I have (currently free but maybe paid in the future) accounts for side projects with businesses that Claude chose for me. (I reviewed the suggestion, but didn’t see a problem or reason to change it.) Why did Claude choose those providers? No idea, but if it’s through subtle marketing to Claude, then well done to them.
Websites cost money. Websites won't cater to bots the same way they don't cater to the scrapers of old. Some bot companies may be charged extra to integrate their bots in some websites, though!
This will go the same way, in 2 years 99% of Internet traffic will be AI agents. You close the doors, you stop existing.
Recently I've asked an agent to find me a laptop with particular requirements. You do not let the agent search your site, you will not get the sale.
For example, I now have to `pacman -S chromium` to buy tickets from Deutche Bahn. They accuse me of being a bot when I use firefox on linux. It's a known problem, and I think stuff like this will only get worse the longer this cat and mouse goes on.
I mean there are multiple reasons not just the AI agents but that's also a part of it
Web 2.0 went from serving users to profit maximization. Amazon became unusable. Streaming services fractured and fragmented, with content locked behind numerous paywalls. Ads permeate every surface.
This is not to say the agentic era will end differently. But at the onset of this era, and for the foreseeable future, customer adoption and trust will decide winners and losers.
Not only this, but the AI will start picking up the scam websites that allow it to search it.
I expect that will no longer be the case in the next year.
(Shrug) Well, they must not want to sell anything to me.
This is not a battle that site owners want to fight. They may think it is, but they're horrifically wrong about that.
You are in a bubble. Normal people (i.e. 99% of people) will not use these agentic tools:
* Some, probably most, won't use them because they don't trust them
* Agentic tool providers can't possibly supply everyone with their own free agentic tooling
* The vast majority of people will not pay providers what it will cost to run these agentic tools (or even anything)
* The vast majority of people cannot afford to buy their own hardware to run their own agentic tools in-house
* Agentic tooling is mostly useless for most people (who cares that it can check you into flights, most people don't have that kind of problem)
For example, higher return rate for online shopping (because the tools didn't buy the correct thing, or maybe they bought something that they shouldn't have).
Or, the tool ignores the adverts on an ad-supported website, intended for humans, basically just adding costs for no benefit to the site.
Or, the tool wastes customer services' time on something that ultimately the agent's operator didn't even want in the first place.
They're just off the top of my head.
So the end result will be that the sites spend money on bandwidth to shove ads and dark patterns at me, which I will never, ever see.
You'd think they'd understand who's going to end up with the power to say "No" here, but apparently not.