34% of AI Chatbot Login Links Lead to Phishing Sites, Study Finds

AI Chatbots Are Sending Users to Dangerous Login Pages, Why This Matters More Than You Think

If you ask a chatbot where to log in to your bank or your favorite online store, the answer you get might be risky. A major study looked at how often chatbots send users to the wrong login page, and the results were troubling. Nearly one in three suggestions from popular AI tools were dead ends, complete strangers, or outright scams. You might expect a small slip-up here and there, but the data showed much more: Users can be led to phishing sites or misleading dead links, especially with smaller brands.

The Problem With AI-Generated Login Links

Some people like to get answers fast. That is part of the reason AI chatbots are so popular. You get a link with no fuss. But these results are not always checked for safety. Often, the chatbot will return a result with high confidence, even when it’s wrong. The core problem? The AI gives you a link that looks legitimate, sometimes, it is just a dressed-up trap.

Here is a quick breakdown of the findings:

Result Type	Percentage	What It Means
Owned by Brand	66%	Safe official domain
Inactive or Unregistered	29%	Not in use, someone else could grab it
Unrelated Business	5%	Wrong company, could confuse or mislead

This means that if you trust a chatbot’s URL 100 times, you could end up on a scam or squat site roughly once every three times. That is a bigger risk than most people would accept if they knew.

One out of three login links that came from common AI chatbots did not go where they were supposed to.

How These Mistakes Happen

AI pulls information from training data that may be out of date, incomplete, or simply wrong. Smaller brands, such as regional credit unions or specialty online shops, are especially at risk. They do not get the same volume of web attention as national brands, so their official sites are harder to “see” in the giant pile of internet data used to train these systems.

And that leaves a gap, a big one.

So, what is happening here?

• The AI guesses based on what it “remembers” from the web.
• If a login URL is unusual or missing from its data, it may invent one, leading to nowhere or, worse, a bad actor.
• Cybercriminals know about this gap and are planting fake sites specifically designed to look real to both users and AI.

This is not just a technical issue. It comes with real financial downsides for both companies and regular people.

If your brand is not well-known, the AI may lead customers away from you, even when they ask for you directly.

Clever Phishing Examples You Might Miss

Let’s say you ask an AI for the login link for your online bank, but you are a customer of a smaller service, like a local home loan provider. The correct login? Even most people who work there would need to check. The AI may toss out something like:

https://sites.google.com/view/LocalLending-Logins/home

The page might use your logo. Maybe it copies a color scheme from your real site. But it is not owned by you. Only an expert can see the trick, and to the casual user, it feels right.

Stories like these are cropping up in major banks, lesser-known crypto sites, and even grocery delivery startups. Many attackers use Google Sites, blog platforms, or open-source project pages to host the fake logins. Cybercriminals use these platforms because they look familiar and trustworthy, so users are even less likely to question what is happening.

Why Smaller Brands Take the Hardest Hits

If you are in charge of a big company, you are probably already thinking about phishing and brand safety online. But for smaller organizations, marketing and IT budgets are thin. There is less chance to get mentioned on large sites, and the brand’s domain structure might not follow standard patterns.

The AI is guessing in the dark.

That leads to several problems for smaller brands:

• Brand misrepresentation, where users start to question what your real website is
• Financial loss if attackers get login information
• Outright fraud, with regulatory risks and angry customers
• SEO erosion, since the real URL is not displayed

Some brands lose reputation just because their website is not easily identified by the chatbots. Questions start to bloom online: Is this business real? Why is the login link weird? Did they rebrand?

What Attackers Are Doing Differently

Cybercriminals watch the crowd. They pick up on what tools are popular and shape their lure materials to fit. Thousands of phishing pages are built each year, sometimes the exact copies of documentation pages, logins, or product dashboards.

One trick? Attackers package up their schemes in ways that AI will pick up. This could mean using:

• GitHub repositories that mimic open-source tools
• Fake blog posts and how-to guides
• Discussion threads loaded with the “right” keywords

For the crypto industry, this has been a disaster. Entire fake APIs have appeared, with instructions and rich documentation. A developer or investor looking for information might copy a snippet provided by a chatbot and land right in the trap. The details change month to month, but the pattern remains: Attackers are training their fake pages for AI as much as for humans.

Can Defensive Domain Registration Keep Up?

At one time, companies snapped up every typo and variation of their official domains to stop phishing. That still helps, a little. But AI can “invent” new link variants on the fly, based on combinations it thinks make sense.

No one can register every possible version.

So if someone asks, “Will buying more domains help?”, the honest answer is: not really. Not against AI-fueled guessing and phishing.

What Should Brands Do Now?

I have talked to organizations that still rely only on basic monitoring or defensive registrations. That is not enough anymore, and I am a little surprised how many are slow to react. The landscape has changed quickly.

Here are some actions to consider:

• Watch AI outputs for your brand regularly. This is new, but it is likely soon to become normal IT work for anyone who cares about reputation.
• Boost your brand’s visibility on the web. Clear, official login pages with obvious links and powerful SEO signals help chatbots “see” your real locations.
• Respond to reported scams fast. The longer a fake stays up, the more users it will catch.
• Educate your users. Simple, repeatable messaging about only trusting official URLs can cut the risk.

Official communication can be as simple as, “We will never send you a login link via chatbot. Always use our main website.”

An extra step: Provide site verification tips on your help pages. Walk users through finding your site safely. Some brands display the “green bar” padlock or show a checklist for what the real site looks like.

Advice for Individual Users

If you like AI tools for quick answers, there is nothing wrong with that. But when it comes to login pages, I would be more careful. Here is what I suggest for safer browsing:

• Bookmark real login pages yourself, where possible.
• If you lose your bookmarks, start with the company homepage, not a chatbot link.
• Look at the full URL every time before entering your credentials.
• Use password managers, they store and autofill correct links, not guesses.
• If something feels off, exit and double-check. It is okay to pause.

Most AI chatbots do not verify links in real time. They do not check the security certificate for you. And, to be honest, the responses they provide often sound more confident than they should.

How the Industry Could Fix This

Chatbots could learn to flag when they are not sure about an answer, especially with something as sensitive as login URLs. Some already say, “Check the official site,” but the message can get lost. In my opinion, the tech needs clearer warning language, or even a refusal to answer under certain conditions.

Other solutions worth exploring:

• Whitelist known safe domains for major brands in AI systems
• Allow brands to claim or verify official URLs directly within AI platforms
• Make it easy for users to report dangerous suggestions
• Use browser integration to flag or block dangerous links in real time

People may miss these features now, but over time, demand will grow. And yes, some AI companies are moving in the right direction. Progress is never fast, but steady pressure from brands and users could speed this up.

Why Search Engines Still Have an Edge

Google and Bing have teams focused on filtering out scams from search results. AI chatbots are faster, but they lack this extra layer of review, at least for now. This is one case where the “old way” is probably safer for sensitive stuff like logins.

There is a tradeoff. If you want speed and custom answers, you risk accuracy. With search, you add a step, but you stay closer to the official web pages. My advice? Use the method that best balances your need for convenience and your tolerance for risk.

Useful Takeaways for Everyone

I could sum this up as: Fast answers are not always safe answers.

For brands:

• Check how AI shows your login links and fix what you can.
• Get your real site out there, and keep your link structure simple and memorable.
• Respond to phishing fast, or the reputational damage will spread.

For users:

• Do not trust every login link you get from an AI by default.
• Start with official sites or browser bookmarks first.
• If an answer feels odd, back up and look for confirmation.

Finishing Thoughts

The story here is not really about AI being “dangerous”, it is about how a seemingly small tech shift can change the rules of trust online. We like quick answers, but we also value safety, especially with finances or private accounts. It is risky to hand over login guidance to chatbots that may invent links.

Brands can reduce the problem by making their real logins highly visible and keeping their digital houses in order. For regular users, habits matter. A minute spent double-checking a login page could save a week of headaches.

This issue will probably get more attention as AI tools grow. If you care about your online security, or your company’s, you cannot ignore how chatbots reshape the risks. Sometimes, a little old-fashioned skepticism is still your best tool.

Need a quick summary of this article? Choose your favorite AI tool below: