Retail trading has a new sales pitch. Where the adverts once promised zero commission, they now promise intelligence: bots that trade while you sleep, algorithms tuned by machine learning, copy-the-AI portfolios with backtests that look like ski slopes. UK search interest in AI trading tools has climbed steadily through 2025 and 2026, and the industry has noticed. The problem is that almost none of the claims are independently checked, and the gap between marketing and reality in this corner of fintech is as wide as anywhere in retail finance.

What the bots actually do

Strip away the branding and most retail AI trading products fall into three buckets. The first is signal generation: software that scans markets and suggests entries and exits, with the human still pulling the trigger. The second is automated execution of rule-based strategies, which is less artificial intelligence than fast arithmetic, however it is marketed. The third and smallest bucket is genuine adaptive automation, systems that adjust their behaviour to changing conditions. The pricing rarely maps to the bucket: some of the most expensive subscriptions sit firmly in the second category, selling decades-old technical analysis with a neural-network sticker on the box.

The backtest trap

Every AI trading product leads with a performance chart, and nearly every one of those charts is a backtest. Backtests are simulations run against historical data, and they flatter for structural reasons: survivorship in the strategy selection, fees and slippage modelled generously or not at all, and the quiet reality that a strategy tuned until it fits the past has learned the past, not the future. The honest test is a live, funded account run forward in time, and vanishingly few vendors publish one.

Testing with real money instead

That verification gap is slowly being closed from outside the industry. Independent reviewers have started running these systems with live deposits and publishing what actually happens. Comparisons of the best AI trading bots available to UK traders now rank tools on forward performance, genuine automation level, costs and FCA-regulated broker compatibility rather than on the vendor’s own backtest, which changes the picture considerably: several heavily advertised products slide down the table once real execution costs are counted, while quieter tools with modest claims hold up better.

The regulatory seam

UK traders should also understand where the regulation actually sits. The bot or signal service itself is typically unregulated software; the protection lives at the venue where the trades execute. Running any automated strategy through an FCA-authorised broker means the account sits inside the compensation and conduct regime even when the software layer does not. Independent rankings of the best CFD trading platforms in the UK increasingly note API and automation support for exactly this reason: the venue choice is the regulated half of an automated setup, and it deserves at least as much scrutiny as the algorithm.

A sensible checklist

For anyone tempted by the AI pitch, four checks separate the plausible from the fantastical. Ask for forward performance, not backtests, and treat refusal as an answer. Price the full stack, subscription plus spreads plus any profit share, against what a simple passive alternative costs. Confirm the execution venue is FCA-authorised and understand what that does and does not protect. And size the allocation as an experiment, not a strategy: if the system is as good as claimed, it will still be good in six months with a larger allocation, and if it is not, the tuition was cheap.

Artificial intelligence will keep reshaping retail trading, and some of the tools are genuinely useful. But in a market where the marketing has outrun the auditing, the trader’s edge is not access to the cleverest algorithm. It is the boring discipline of demanding proof before deposit, which no amount of machine learning has yet managed to replace.