Using AI to write complaint letters? Here’s why it doesn’t work

If you’re thinking of using artificial intelligence to craft a complaint to an airline, cruise line, or car rental company, you may want to reconsider.

As a consumer advocate and TPG’s ombudsman columnist, I’ve observed a distinct change in the help requests that cross my desk. Troubled travelers are using AI to navigate customer service problems — and the results aren’t pretty.

At first glance, the letters read like well-written arguments penned by determined travelers who have done extensive research. They come complete with bulleted lists of “evidence.”

Unfortunately, they can also sound like unnecessarily aggressive legal briefs rather than genuine pleas for help. What’s more, the laws and regulations cited, not to mention their basic premises, are sometimes AI hallucinations, meaning they do not exist. Nearly always, the complaints conclude with highly inflated compensation demands, accusations of fraud and hard deadlines for a response.

Unsurprisingly, companies often ignore AI-generated complaints that are filled with legal threats and have little basis in reality. Even worse, the outcome for some travelers is a bigger problem than the original issue. Trusting the advice of an AI agent can backfire spectacularly, even putting the customer at risk of landing on a travel company’s blacklist.

The travelers you’re about to read about learned the hard way that not everything an AI agent spits out is true. Here are their stories and what you should know before you sign your name to an AI-spun complaint.

AI interprets a cruise passenger’s buyer’s remorse as fraud

Cruise passengers often contact me after they’ve made purchases on board a ship or in ports of call that they regret later. The items are typically nonrefundable and nonreturnable and come with a giant price tag.

Things like jewelry, artwork, beauty products, timeshares and more that were a must-have while under the influence of what I call vacation brain don’t look quite as appealing at home. The traveler, suffering from extreme buyer’s remorse, hopes I can negotiate the return of an expensive, unwanted souvenir.

But buyer’s remorse is not the basis for breaking a signed contract, which is what a receipt amounts to. So when a traveler asks for this type of help, I am forced to explain the reality of their situation. For most, it’s a hard lesson about making spontaneous, high-ticket purchases during vacation with no refund option.

That’s what happens when someone with buyer’s remorse reaches out to me, since I am a real human being who is familiar with real rules and regulations. However, I recently discovered what happens when a cruise passenger asks AI bots the same question.

Earlier this year, Amanda L., sailed on Royal Caribbean’s Star of the Seas. On the first night of the cruise, she bought a $25,000 ring in the ship’s Effy jewelry store. She made no complaints about the purchase during the cruise, but buyer’s remorse settled in the moment she disembarked.

That evening, as she was looking at her new bauble, she began to think it was overpriced. It didn’t look so dazzling now that vacation was over.

She asked her AI program of choice, Claude, its “opinion” of the situation. Amanda shared the documents she had received at the boutique, including the appraisal by Effy’s certified gemologist, a photo of the ring and an account of her recollections of conversations with the salespeople.

Instantaneously, AI had a reassuring answer for Amanda: She had paid too much for the ring. Not only that, but because of how she framed her question to Claude, it had determined that she was a victim of “predatory sales practices” and fraud. When she asked what she should do, AI advised her to immediately file a credit card dispute despite the fact that the $25,000 ring that she had willingly purchased was sitting on the table in front of her.

Soon, she was on the phone with her credit card company repeating to the representatives the same thing AI told her. The big-dollar charge was put on hold that night while her credit card company investigated.

But Amanda’s brainstorming session with her AI wasn’t over just yet. While her bags weren’t even unpacked from the cruise, Claude told her that she should file a fraud report with the Federal Trade Commission claiming that the Effy salespeople had told her the ring was a great investment, but in reality it was overpriced and overvalued.

Next, AI crafted a strongly worded letter for her to send to the legal department and president of Royal Caribbean. It also recommended that she file a formal complaint with the Better Business Bureau accusing Effy of fraudulent business practices.

The AI instantly wrote that complaint, too.

It had been a busy day for Amanda and her AI companion.

“Formal complaint — predatory sales practices and falsified documentation.”

AI had convinced Amanda so completely that she wasn’t simply suffering from buyer’s remorse that it led her to make legal demands and threats before trying to resolve her problem directly with Effy. When a representative of the jewelry store called and emailed her in an attempt to resolve her complaint cordially, she told him that Effy would be hearing from her lawyer.

More than two months after Amanda began chatting with Claude about her jewelry purchase, she was no closer to getting the ring accepted for a return and the charge refunded. In fact, all the companies involved had stopped responding to her. That’s likely because the inflammatory AI-written letter to the president and legal department of the cruise line accused Effy of falsifying documents, fraud and predatory sales practices and suggested Royal Caribbean condoned it all.

Frustrated by her lack of progress, she again turned to Claude and asked what else she could do.

That’s when AI told her it had found an article I had written about another Royal Caribbean passenger who thought his Effy jewelry purchase was wildly overpriced. Claude suggested that since I had been able to help that traveler, I might be able to help her, too.

Soon, Amanda’s request for help hit my inbox. By then she had lost her credit card dispute, which had virtually no chance of succeeding in the first place since Amanda had signed the receipts for the ring and left the ship with it, making no complaints until she arrived home.

The Fair Credit Billing Act allows credit card users to dispute billing errors, including unauthorized charges. But Amanda’s purchase of the ring was neither a billing error nor unauthorized.

Royal Caribbean provided her credit card company with the signed receipts to defend against the chargeback. And despite the AI bot’s allegation that the jewelry store had committed fraud, its bulleted list of evidence supporting that accusation included minor, inconsequential details, such as the wrong salesperson listed on the receipt and a duplicate appraisal having the wrong date.

What the paper trail actually suggested was a cruise passenger regretting the impulse purchase of an expensive ring. That’s not exactly a stop-the-presses moment. The price of jewelry on board cruise ships is often inflated. That’s not fraud, that’s just marketing and sales tactics.

But when I suggested that Amanda retract her fraud complaint with the FTC, Amanda turned to Claude again. Shortly after, I received an email with the subject line: “Why this isn’t buyer’s remorse.” It contained all the hallmarks of an AI-written rebuttal in a well-organized sequence with headers and bullet points. But it repeated the same information it had already previously stated, and in some places, it actually contradicted itself. At the bottom, Amanda signed her name.

In my reply, I got right to the point. “I’m concerned that you are relying on AI to navigate this problem,” I told her. “Your AI is telling you what you want to hear rather than what you should know. AI programs are highly susceptible to suggestions of the user.”

Pointing out some of its most blatant mistakes, I suggested that Claude had caused Amanda to escalate her complaint to the nuclear level while completely bypassing basic problem-solving steps.

“Yes, Claude makes mistakes sometimes,” Amanda wrote back. “But for someone like me who has never experienced this type of thing before, he has been very calming and helpful.”

“He” had also caused Amanda to miss the 30-day window of opportunity when Effy would have accepted the return of the ring. While Claude was sending Amanda on a path to the legal department and the FTC, her problem could have been easily resolved by contacting Effy and asking for a refund.

Asking Royal Caribbean if there was any chance of a jewelry return

Effy offers a 30-day return policy on the jewelry sold in its shops, including the ones purchased on board cruise ships. Effy confirmed to me that had Amanda simply contacted customer service on the day AI encouraged her to make fraud complaints with the FTC, BBB, Amex and Royal Caribbean, she could have returned the ring for a full refund. Within 60 days of the purchase, she could have returned the ring for an in-store credit.

Unfortunately, during that entire 60 days she was following the aggressive advice of her AI agent. By the time her AI bot suggested she contact me, the return window was past and a goodwill gesture seemed unlikely given the steps Claude had advised her to take.

I still sent a request over to my executive contacts at Royal Caribbean to see if they might overlook all those complaints and the extended return delay. Unfortunately, it seems the damage was done.

Royal Caribbean declined to comment on Amanda’s case. That isn’t unusual. After a consumer sends accusations to a company’s legal department, that generally shuts down cordial communication and the possibility of a goodwill gesture. Today, Amanda still has the ring and although a local appraiser in Florida valued it at about half the price she paid for it, it looks like she’s stuck with it.

Even worse, I suspect that Amanda may no longer be welcome aboard any Royal Caribbean ship. I’ve shared this information before, but cruise lines are very willing to permanently ban passengers after they make brand-damaging or excessive complaints.

Not alone in following bad AI advice

Amanda’s complaint may be one of the most spectacular examples I’ve seen of AI-driven problem-solving gone wrong, but she is far from alone.

Car rental customers and hotel guests have also arrived at my helpline armed with AI-generated complaints, overly confident allegations and “critical evidence” that doesn’t withstand even basic scrutiny.

Malky Y. reached out to me after Sixt accused her of damaging a rental car and charged her $2,300. She insisted she hadn’t damaged the vehicle, but had no before-and-after photos needed to prove it.

This is a common problem. Unfortunately, car rental customers often get hit with repair charges for damages they are sure they didn’t cause. That’s why I created free car rental inspection checklists travelers can use to protect themselves, documenting a vehicle at pickup and return.

When Malky received a batch of pictures from Sixt, she uploaded them to AI for analysis. Within seconds, the program announced it had found something “very significant” in her favor.

“CRITICAL FINDING: ALL PHOTOS SHOW THE IDENTICAL TIMESTAMP,” reported her AI agent, which also happened to be Claude, concluding that Sixt had reused or possibly falsified evidence. Claude advised Malky to accuse the company of timestamp fraud, dispute the charge with her credit card company and file a complaint with the Florida attorney general.

There was just one glaring problem: When I reviewed the photos, I could see they were all taken as Malky exited the lot at the beginning of her rental. Of course they had the same timestamp.

The photos she needed were the ones taken when she returned the car. Although Sixt had sent the AI scanner’s full report which contained the before and after scans, the link had expired and she had only downloaded the before photos. So that’s what she shared with Claude.

AI had misidentified various angles of the same scan as before-and-after evidence rather than just the before photos. Then, building on its own mistake, it devised an entire legal strategy and assured Malky she had a “strong case.”

Luckily for her, she hadn’t sent the accusatory letter to Sixt. Claude had also recommended that she send her complaint to me for review and I’m currently working with Sixt to review the actual before and after photos.

Another traveler, Emilia G., used AI to help translate and prepare a complaint about a Booking.com apartment rental in Italy that she feared didn’t exist. Her underlying concern was legitimate. The host was unresponsive, a suspicious payment request had appeared in Booking’s messaging system and her family was scheduled to arrive within a few days.

But AI transformed those facts into accusations that Booking was “knowingly” sending her family into the hands of scammers, breaching European law and intenting to leave Emilia’s family stranded on the street. It also created a bizarre illustration depicting the company operating a deliberate “complaint shredder” designed to extinguish claims and avoid responsibility.

Her AI titled the complaint, “Phishing, fake numbers, and breached EU law. Booking.com’s massive failure right before a family of 5’s flight.” Then it encouraged her to send it as a press release blast to every media outlet possible. The only thing is, at the same time, Booking was working to figure out whether or not the property was indeed a problem.

But the AI bot’s provocative complaint buried Emilia’s urgent problem under a mountain of speculation to anyone who received it.

After I read through it, I asked Emilia to put aside the AI-generated narrative and send me only her reservation and correspondence with Booking. The paper trail actually showed that Booking had already removed the property from its listings, was processing her refund and had offered additional money toward replacement accommodations.

Again, AI had escalated a problem instead of attempting to resolve an issue at the most basic level.

Both travelers may have had legitimate reasons to challenge the companies involved. But AI didn’t investigate their complaints. It took their suspicions, presented them as facts and then manufactured evidence and legal arguments to support them.

How to use AI responsibly when writing a complaint

AI can be a useful tool for organizing your thoughts, improving grammar, or making a complaint more concise. But it shouldn’t be trusted to investigate your problem, interpret the law, or decide that you’re a victim of fraud.

If you choose to use AI to help write a complaint, keep the following in mind.

Remember that AI makes mistakes

AI programs often deliver incorrect information with complete confidence. It can be very convincing, especially if it is saying what you want to hear. That polished, authoritative answer may contain laws that don’t exist, regulations that don’t apply and conclusions unsupported by the facts.

Never assume something is true simply because AI presents it as fact — even if the answer includes official-sounding language and citations.

Ask AI to challenge your position

Many consumers explain their problem to AI and ask it to confirm why they’re right. That can lead the program to build an argument around a flawed premise.

Instead, ask it to identify weaknesses in your position. You can also ask what facts might support the company’s side or what information you may be overlooking. That’s much more useful than asking AI to validate the outcome you already want.

Don’t let AI create a legal strategy

Unless you’re working with a licensed attorney, don’t allow an AI program to turn an ordinary customer service problem into a legal battle.

Accusations of fraud, discrimination, criminal behavior, or nefarious intent can immediately end a company’s willingness to communicate informally. Threatening lawsuits, regulators, or the media in your opening complaint will rarely encourage anyone to offer a goodwill resolution, either.

Start at the lowest reasonable level and give the company a fair opportunity to fix the problem.

Fact-check every claim

Before putting your name on an AI-generated complaint, verify every law, regulation, policy and quote it contains. Click through to the original source — not another AI summary of it — and make certain that the information applies to your specific circumstances.

If you can’t confirm a claim independently, remove it. Don’t forget to use your common sense.

Make the final complaint your own

Use AI as an editor, not as your investigator, attorney, or spokesperson. Your complaint should clearly explain what happened, what evidence you have and what reasonable resolution you’re requesting. Do not let it convince you to make exorbitant recompense demands.

Most importantly, read the finished letter carefully before sending it. Get a second opinion and ask a friend to read it, too.

Bottom line

AI can help you organize your thoughts and refine a complaint, but it can’t replace common sense — or a real human being with expertise in the problem you’re trying to solve.

Never allow an AI bot to steer your complaint, invent a legal strategy, or make serious accusations on your behalf. Verify everything, keep your demands reasonable and give the company a fair chance to fix the problem before escalating.

And remember: Once you put your name on an AI-written complaint, every claim in it belongs to you — including the ones your digital assistant made up.

If you have a problem with an airline, car rental company, hotel, cruise line, or credit card company and want a real consumer advocate’s assistance, send your request to ombudsman@thepointsguy.com and I’ll try to help. But do it before an AI agent gets you banned from your favorite company.

Disclaimer : This story is auto aggregated by a computer programme and has not been created or edited by DOWNTHENEWS. Publisher: thepointsguy.com