Complaints about online platforms often go nowhere. AI could actually be part of the solution

You know the feeling. Your account is suspended, your refund request is refused, or you’re still being charged for a cancelled subscription.

So you go to the company’s website for help and you find a chatbot. It offers a menu that doesn’t include your situation. There is no option to speak to a person. Eventually you give up.

Since the start of 2023, Australians have made 1,780 complaints about online retailers, marketplaces, search engines and social media platforms to the telecommunications industry ombudsman. In 2025, complaints were up 20% on the year before.

Account access problems were the most common problem, followed by disputes over charges and fees. More than seven in ten involved just five companies: Meta, Google, Microsoft, Apple and Hubbl.

Unfortunately, all those complaints about online services made to the telecoms industry ombudsman fall outside its jurisdiction. No existing ombuds scheme can make a binding decision about them.

We need a dispute resolution system that can deal with these complaints, and there are precedents for this.

Digital problems cost the economy

The Consumer Policy Research Centre found four in five Australians had a problem with a digital service in the past year, and that unresolved problems cost an estimated $497 million annually. Two in three who complained were left dissatisfied.

Dissatisfaction was highest over refunds people believed they were entitled to (84%), followed by having no way to contact someone who could help (83%).

The centre says when people complained to online businesses and platforms, many had to deal with virtual customer service portals and AI chatbots that are unhelpful, give no explanation or offer inaccurate advice.

Who can deal with these complaints?

The government’s scams prevention framework offers a model. It requires banks, telcos and global digital platforms to join an external
dispute resolution scheme run by the Australian Financial Complaints Authority by September 1.

However, this scheme is limited to scams and won’t cover the sort of account or platform issues outlined above.

Under Europe’s Digital Services Act, platforms must have a proper complaints process, and decisions cannot be made by machines alone. Qualified staff must supervise them. If the platform still gets it wrong, users can take the dispute to an independent body certified by the government, which must decide within 90 days.

For example, one dispute settlement body, Appeals Centre Europe, has ruled on 1,500 disputes and overturned more than three quarters of the platforms’ original decisions.

Maybe AI could be part of the solution

Any dispute resolution scheme covering online platforms would face huge numbers of cases with similar problems. This is where chatbots and AI could actually become part of the solution.

British Columbia’s Civil Resolution Tribunal guides people through an interactive Solution Explorer, a rules-based chatbot that asks set questions and returns tailored information based on set rules, before any human decision-maker gets involved.

In research my colleagues and I published in the UNSW Law Journal, we compared rules-based chatbots with generative AI chatbots to see how they can help people with small-value claims.

Generative AI’s ability to provide personalised and targeted responses is genuinely attractive.

But generative AI also hallucinates, confidently inventing legal rules and cases that don’t exist. Even when it isn’t fabricating a source, it can
produce highly convincing advice that is simply wrong.

That risk falls hardest on the people these systems are meant to help. A consumer who can’t afford a lawyer is also the person least equipped to notice when an AI has made up the law. And the error is invisible: told wrongly that you have no claim, you simply walk away, and nobody ever finds out. No feedback reaches the system that produced the error.

Getting the design right

The Victorian Law Reform Commission sets out sensible principles in its report on AI in Victoria’s courts, including one clear recommendation that AI must not make the final decision. A person must decide the outcome.

Before any new resolution scheme integrates AI tools, it must decide which tools will be used, for what functions, and what safeguards will reduce potential risks.

The chatbots people are stuck in today aren’t broken. They are working exactly as designed – for the online companies that built them.

Australia rarely gets to design a dispute resolution scheme from scratch, and it is building one now as part of its scams prevention framework.

Whether technology ends up serving consumers or stonewalling them is not a question about AI. It’s a question about who the system is designed for, and whether there is always, at the end of it, someone to turn to.

Vivi Tan does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.