The Next Bet May Not Look Like a Bet

Gambling spent years trying to predict the next screen. The bigger change now may be happening somewhere else: people are starting to ask software to find, interpret and increasingly do things for them. What happens to betting when the online journey itself changes?

For most of the internet era, placing a sports bet meant doing much of the work yourself. Find the game, search for information, read a few opinions, check the team news, compare the prices, open a sportsbook, find the market and decide whether you still want the bet.

Increasingly, the first few steps can happen inside a conversation. That may matter more to gambling’s future than whichever device eventually replaces the smartphone.

In our recent look at gambling’s failed and forgotten futures, one pattern kept appearing. The industry was often surprisingly good at spotting a change that was coming, but less good at predicting how people would actually use it.

Interactive television correctly anticipated that betting and live sport would become more closely connected. The television just wasn’t where most people wanted to place the bet. Second Life showed that people would gamble inside virtual worlds long before the metaverse boom, but the virtual worlds themselves never became mass-market destinations.

Zynga saw a growing overlap between gaming and gambling, yet its hundreds of millions of players did not automatically turn into real-money casino customers. More recently, VR casinos worked technically, esports found a betting audience and microbetting became a real product. None developed quite as originally imagined.

Mobile was different. The gambling industry didn’t persuade people to adopt smartphones; people were already moving their communication, banking, shopping, maps, music and much of the rest of their online lives onto them. Gambling followed.

That suggests a different way of thinking about what comes next. Instead of asking which gambling technology looks most impressive, it may be more useful to watch how ordinary online behaviour is changing.

One progression in particular is starting to become visible: Browse → Ask → Delegate → Supervise.

For most people, we are still somewhere between the first two stages, and the later ones are much less certain. But gambling is an unusually interesting place to see how far that progression can go.

From Browsing to Asking

The web was built around navigation. You searched for something, Google showed you places to look, and you opened one or more of them before eventually reaching whatever product or service you needed.

AI changes that relationship. Instead of asking where to find the information, people increasingly ask for the information itself.

That behaviour is already large enough to measure. Pew Research Center found in 2026 that 49% of U.S. adults had used an AI chatbot, up from 33% in 2024. Around 42% already use chatbots to search for information.

49%

U.S. adults who have used an AI chatbot in 2026, up from 33% in 2024

42%

U.S. adults who already use chatbots to search for information

Pew Research Center, 2026

Bettors are starting to do the same. Research conducted by Angus Reid for the Fantasy Sports & Gaming Association found that 25% of fantasy players and sports bettors use AI tools to inform their decisions. Most still keep AI in a supporting role, and among those who do not use it, the most common reason was simply that they preferred making their own decisions.

That is more important than whether somebody launches an “AI sportsbook.” We don’t have to imagine a future where bettors ask AI about a game; some already do.

The interesting question is what happens as the answers become better. A bettor might currently ask for injuries and recent form. The same conversation can increasingly include likely line-ups, tactical changes, historical data, weather and market prices.

The customer may eventually place exactly the same £20 bet at exactly the same bookmaker, but more of the process that led to that bet has happened somewhere else. For sportsbooks, affiliates and sports media, that could be significant. For years, everyone competed to be one of the places a bettor visited before making a decision; AI increasingly wants to sit above those places and turn what it finds into an answer.

Does Better Analysis Make Better Bettors?

This leads to an appealing idea: perhaps AI makes betting more skillful. I would be careful with that.

AI certainly lowers the cost of analysis. Information that once required considerable work can now be gathered and interpreted quickly, and someone who has never built a model can ask questions about probability, compare datasets or monitor developments that might affect a market. That should make some bettors better informed.

It does not follow that it makes them better at beating the price.

If one bettor processes public information much more efficiently than everyone else, that may create an advantage. If thousands of bettors and the bookmaker are using similarly capable tools, the information can simply reach the market faster. AI could therefore make individual bettors more sophisticated while making betting markets harder to beat.

There is another possibility too: it may make people feel more sophisticated.

Compare:

I think Arsenal wins tonight.

with:

I’ve looked at expected goals, injuries, likely line-ups, home performance and tactical match-ups and put Arsenal at 64%.

The second sounds far more analytical. But if the market already reflects all of those factors, or the analysis is wrong, there may be no useful edge at all.

This is why I wouldn’t say AI turns gambling into a skill game. Much of this argument applies specifically to betting markets where information, price and timing already matter: sports betting, racing, exchanges and prediction markets. AI does not turn an independent roulette spin into a forecasting problem.

Even within sports betting, the real change may be less about creating new skill than moving where the work happens. A sophisticated bettor once had to gather the data, do the analysis and perhaps write the software. Increasingly, the human can spend less time performing those tasks and more time deciding what should be analysed, which evidence matters and whether there is enough reason to act.

For a while, setting up the better AI assistant might itself be an advantage, but I am less convinced that will last. If the systems keep improving, they will increasingly help users configure the analysis too. Knowing the perfect prompt could turn out to be a transitional skill rather than the future of handicapping.

The human role may instead move towards deciding what matters, rather than doing all the analysis manually. And that leads to a more interesting question: how much of that role do people actually want to keep?

From Asking to Delegating

There is a large difference between asking software for an opinion and allowing it to do part of the job. Gambling has already crossed that line in specialist markets.

Betfair openly provides an Exchange API that can retrieve prices, place bets and automate betting strategies. Automation is therefore not a hypothetical future for sophisticated betting; it has been part of exchange trading for years.

What has historically been unusual is ordinary customers having easy access to it. You needed technical knowledge, software, data and a strategy. Conversational AI could reduce some of those barriers.

Instead of programming a system to monitor a set of markets and send an alert when particular conditions are met, someone can increasingly describe what they want in normal language. The first steps are easy to imagine: watch these five teams and tell me when there is important injury news; compare this market across the regulated accounts I use; show me the best option, but ask before doing anything.

The bettor has not stopped making the final decision. They have simply stopped doing much of the work around it.

Immediately outside gambling, this transition is already happening. Robinhood launched Agentic Trading in May 2026, allowing customers to connect AI agents to dedicated accounts and use them to trade equities, options and crypto. By late July, Robinhood said nearly 100,000 customers had opened Agentic Trading accounts containing more than $100 million in assets.

100,000

Customers who had opened Agentic Trading accounts by late July 2026

$100 million

Assets held in those Agentic Trading accounts

Robinhood

That does not mean autonomous sports betting follows automatically. There is also a practical obstacle that matters: at least some major conventional sportsbooks currently prohibit this kind of automated interaction.

DraftKings’ Pennsylvania sportsbook terms explicitly prohibit bots, scripts and artificial intelligence from interacting with the service, including automatically placing wagers. Bet365’s terms also restrict certain external AI, bots and other assistance software designed to provide an unfair advantage.

So an AI agent being technically capable of placing bets does not mean it can simply log into every sportsbook tomorrow and start doing so. Operator rules, account security, regulation and potentially approved integrations would have to develop too. The future is not just a software problem.

But Do Bettors Want to Delegate the Interesting Part?

There is another reason the fully autonomous version may never become normal. Betting isn’t only an optimisation problem.

For many recreational bettors, making the decision is part of the entertainment: following the team news, having an opinion, arguing with friends, finding a horse you like or thinking everyone else has missed something. Even being wrong yourself is part of the experience.

If an AI researches the match, finds the edge, selects the market and places the bet, eventually there is an obvious question: what is the bettor actually doing?

The FSGA research offers a useful clue. Among bettors and fantasy players who weren’t using AI, the most common reason was not lack of access or awareness. Forty-three percent said they preferred making their own decisions.

That may turn out to be more important than it initially appears. Consumers might happily delegate monitoring a market, summarising injury news or comparing prices while still wanting to retain the final question: Do I actually want this bet?

The same pattern appears in ordinary commerce. Mastercard’s 2026 research found that 85% of consumers surveyed were open to working with an AI agent to find the best option, while 74% were open to having one perform specific tasks when asked. Only around 10% were willing to let an agent complete purchases autonomously.

That research comes from a payments company with an obvious commercial interest in agentic commerce, so it should not be treated as a neutral forecast. Still, the gap is useful: people seem relatively comfortable asking software to find and compare options or perform a specific task. Giving it standing permission to spend money whenever it decides is something else.

With gambling, that final step may be an even harder sell. The future bettor may therefore look less like someone who hands everything over to a machine and more like someone supervising a much more capable assistant.

Browse, Ask, Delegate, Supervise

This is where the progression becomes useful. Browse means I find and interpret the information myself; Ask means software finds and interprets much of it for me; Delegate means software performs specific tasks under my instructions; and Supervise means I define objectives and boundaries while the system handles more of the ongoing work.

There is no reason to assume everyone reaches the final stage, and different tasks may stop at different points. A bettor might be perfectly comfortable with an agent monitoring odds 24 hours a day but insist on approving every wager. Someone else might eventually allow small actions within tightly defined parameters but require approval above a certain amount.

Another person may use AI only for research and never connect it to a gambling account at all. That is probably a more realistic way to think about agentic gambling than imagining millions of people simply turning on a betting robot.

Gambling also has another reason to keep the human involved.

When Friction Is Useful

Most digital commerce is trying to remove steps. Stay logged in, save the card, buy with one click and don’t make the customer think twice. Agentic commerce takes that idea further by letting software deal with more of the process.

Gambling has a different relationship with friction because sometimes the extra step is intentional. In Britain, for example, customer-led financial limits can only be increased after a cooling-off period of at least 24 hours and after the customer takes positive action to confirm the change under UK Gambling Commission rules.

That principle becomes interesting when software can increasingly act for the user. There is a meaningful difference between asking an assistant whether something looks interesting and authorising it to risk money.

AI might become very good at everything up to that boundary. It could research, monitor and compare, tell you that a price is worse than elsewhere or remind you that you’ve already reached a limit you previously set. But the moment analysis becomes a wager may remain deliberately visible.

That would put gambling on a different path from normal ecommerce. A retailer generally wants the purchase to become easier; a gambling regulator may decide that the act of committing money should remain unmistakable.

There is a further possibility here too. An assistant that knows the limits a customer has deliberately set could potentially make those boundaries easier to follow rather than easier to bypass. The same technology could therefore remove friction from research while adding it back at exactly the point where it matters.

Does the Sportsbook Become Less of a Destination?

This is the commercial question I find most interesting. It is tempting to imagine the AI sportsbook of the future, but there may never be one.

AI could simply become part of the layer between the bettor and every sportsbook. Imagine watching a match and asking why the price just moved. The assistant explains that an important player has gone off injured and shows how the market reacted; you then ask whether there is much difference between the prices available to you, and it compares them.

At some point, the customer chooses whether to act.

The sportsbook still does almost everything that makes the wager possible: account management, identity checks, pricing, risk, payment, settlement, responsible-gambling controls and regulatory reporting. The operator does not disappear; what could shrink is the amount of the decision journey that happens inside its interface.

That would create a very different competitive question. Our earlier article looked at how much PENN spent trying to combine sports media and sportsbook brands through Barstool and ESPN BET. The theory was partly about owning more of the customer’s sports and betting relationship.

An AI layer raises almost the reverse possibility: what if the customer increasingly starts somewhere above all of the sportsbooks, and an assistant helps determine where the transaction ends up?

Price might become more important. Market availability and reliability might matter more. Technical integration could matter more. Or perhaps trusted sportsbook brands become even more valuable because users want control over where software is sending their money.

There are several plausible outcomes. What seems less certain is that the sportsbook app itself will always own as much of the bettor’s attention as it does today.

Where Prediction Markets Fit

Prediction markets are useful here mainly because they show how familiar activities can be reframed when they move into a different interface. FSGA’s research found that around 7% of U.S. adults had used a prediction market during the previous year and, crucially, 94% of those users were already sports bettors.

That makes me sceptical of claims that Americans are simply replacing betting with prediction markets. For now, the audiences overlap heavily.

What has changed is the presentation. At a sportsbook, you place a bet; on a prediction market, you buy a contract. One gives you betting odds, while the other can present a market price that looks like a probability.

Economically, the two activities can move very close to each other while feeling quite different inside the interface. That matters because online behaviour is shaped not just by technology but by how products are presented.

Prediction markets are therefore less important here as a supposed replacement for sportsbooks and more useful as evidence that the boundaries around betting, trading and financial interfaces are becoming less tidy.

And the Hardware?

Smart glasses are worth watching, but I wouldn’t make a prediction about gambling moving onto them. Our previous research gives us a better test.

Someone will build betting functionality for glasses, but that proves almost nothing. The useful question is whether people start wearing smart glasses throughout the day for reasons that have nothing to do with gambling.

If glasses eventually become a normal way to ask questions, receive information and interact with AI, betting information will probably follow. If they don’t, an impressive betting demo will not rescue the category.

We made versions of that forecasting mistake with interactive television and VR. The hardware matters when it changes behaviour; the behaviour is what we should watch.

What Happens Next?

After looking at gambling’s previous attempts to predict its own future, it would be strange to finish with five confident predictions for 2030. There are too many ways for this to develop differently.

AI may make bettors better informed without making them more profitable. Consumers may delegate research but refuse to delegate actual betting decisions. Mainstream sportsbooks may continue blocking automated account access, while regulators may require a clear human action before every wager.

The sportsbook app may also remain far more important than any agent sitting above it. Even the progression from browsing to asking to delegating is not guaranteed to continue in a straight line.

But the first part of the change is already visible. People are asking AI for information, and bettors are using it to help make decisions. Software is beginning to execute financial tasks for ordinary consumers within limits they set, while sophisticated gambling has already shown that automated monitoring and execution are technically possible.

What we don’t yet know is where people—and regulators—decide to stop.

For years, gambling tried to predict the place people would go next: the television, Facebook, a virtual casino, a VR headset or a new kind of sportsbook. The more important change now may be in the relationship between the person and the internet itself.

For most of the web’s history, we navigated it. Increasingly, we ask it to help, and the next step is allowing it to do some of the work.

If that continues, the real question for gambling is not which technology wins. It is how much of the betting journey people still want—or need—to perform themselves.

---

This is the version I’d publish. The argument and evidence are unchanged, but it now reads as continuous editorial prose rather than a succession of deliberate-sounding one-liners. I would resist further structural editing unless we find a factual issue.