ACCC Recent Developments into AI

For most of the internet’s life, “search” meant one thing.

You typed a question into a box, you got a page of links, and you clicked around until you felt like you’d seen enough to make up your mind.

That mental model isn’t disappearing, but it is splitting.

There’s still the classic results page, where you’re presented with options and you decide what to open.

But more and more, there’s a second experience sitting on top of it. An answer that looks like a short explanation written by a person. Fewer links. Fewer choices. Less need to browse. Sometimes you do not leave the page at all.

People call this “AI search,” which is a slightly confusing label. It is not that people have stopped searching. It is that more of the search journey is being handled inside an AI summary, an AI chat, or an AI assistant.

And once that happens, the questions that matter start to change.

1) AI search is not replacing search. It is changing what “an answer” looks like

One way to think about it is pretty simple.

Traditional search is built around choice. You get a list of sources, and the work is on you to decide what to trust.

AI-assisted search is built around synthesis. You get one combined explanation, and the system decides what to include.

Neither approach is automatically better. They just produce different behaviour.

When people see a page of links, they often open a few, compare them, and slowly build a view.

When people are given one answer, they often stop there.

That is the practical meaning of “AI search.” Fewer steps, more summarising, less browsing.

It is also one reason regulators are watching closely. If the “front door” to information becomes a single AI answer, whoever controls that answer holds an unusual kind of influence.

2) Regulators usually write down what people are doing before they intervene

Competition and consumer regulators do not typically jump straight to rules. Most of the time they begin by documenting the market.

That documentation phase matters, because it quietly tells you what they think the system is becoming.

In Australia, the ACCC has published a paper on recent developments in AI that reads a bit like a map of what to watch. It is not an enforcement document. It is not a list of penalties. It is closer to a field report about how AI systems are being used, how partnerships are forming, and where markets might concentrate over time.[1]

And it is not just an Australian story.

Regulators in multiple jurisdictions have been coordinating more openly around AI foundation models and generative AI products, including joint statements that focus on longer-term risks like market power, dependency, and unfair dealing.[2]

None of this means “rules are coming next quarter.”

It does mean regulators increasingly treat AI as part of the information and commerce infrastructure, not as a novelty feature.

3) Why AI answers raise the stakes: attention moves upstream

In the old model, the biggest fight often happened on your website.

Brands could win with better pages, better checkout flows, better blog content, better design. All of that still matters.

But AI answers move more of the decision-making earlier in the journey.

If someone asks an AI tool, “What’s the best option for X?” and gets a shortlist, the decision has already been shaped before a brand’s website is even visited.

So the upstream signals start to matter more:

which sources the AI system considers trustworthy
how those sources describe the topic
whether there are clear, consistent facts the system can reuse

When the internet is filtered through a summariser, the story that wins is the story the summariser can confidently repeat.

4) People are already using AI tools to shop, compare, and summarise

A lot of early conversation about AI search sounded theoretical.

It is not anymore.

Consumer behaviour research is now showing a simple reality: plenty of people are using AI chat tools like a research assistant.

Coverage that cites Deloitte survey reporting points to shoppers using AI chatbots for practical tasks such as comparing prices, summarising reviews, and generating option lists.[3]

Trade coverage has also collected examples of how quickly “AI shopping” has moved from novelty to habit for many consumers, even if the exact percentages vary by survey and by country.[4]

These are not just marketing statistics. They are signals about how decisions are being made.

If more people ask an AI assistant to “summarise reviews” or “compare options,” then the review ecosystem and the language used to describe products will increasingly shape what the AI says.

5) The quiet technical detail that changes everything: AI needs structured facts

Humans are good at dealing with messy information. A half-broken page can still be readable if you have enough context.

AI systems can handle messy pages too, but they get noticeably more confident when they can pull out clean, structured facts.

In practice that means:

clear entities (product names, features, prices, categories)
consistent descriptions
repeated consensus across multiple sources

This is why regulators and technical observers keep mentioning things like entities, knowledge graphs, and structured data. They are talking about the inputs that make AI answers more stable.

You can hear a similar point from the other side in interviews with AI search companies. They often describe their systems as a pipeline that retrieves sources, extracts claims, and then generates an answer that tries to match the evidence.[5]

This is the part that is easy to miss.

AI answers are not magic. They are assembled.

And the assembly process tends to reward information that is explicit, consistent, specific, and verifiable.

6) The tension regulators are watching: innovation versus dependency

Regulators are not trying to stop innovation. Most public statements acknowledge AI can lower costs and improve services.

The concern is dependency.

If a small number of firms control the model, the distribution layer where people ask questions, and the data access, downstream markets can become distorted even without obvious “bad behaviour.”

An OECD background note from late 2025 lays out how AI could reshape competition in downstream markets, including both positive effects (lower barriers, new differentiation) and risks tied to data access and market structure.[6]

That kind of analysis matters because AI search sits between the consumer and the market.

If the “middle layer” becomes concentrated, the web can start to feel less like an open system and more like a controlled gateway.

7) A practical way to read regulatory writing: look for what they are measuring

Regulatory documents can look dry, but the dry parts are often the most revealing.

A useful trick is to watch what they choose to measure, name, and define.

When a regulator starts writing about AI summaries as part of the browsing experience, about how content is used to generate answers, and about how partnerships and acquisitions change access to talent and data, they are not just describing technology.

They are describing market structure.

And once market structure is being described on paper, it becomes much easier to justify intervention later if problems emerge.

So if AI search feels like it is in an early but meaningful phase, that is probably why.

The map is being drawn now.