Trade Ideas' Holly AI: what the reviews agree on, where they contradict, and what nobody has audited

Most 'AI trading' products today are chatbots. Holly is older and different: a strategy selector. Trade Ideas' learning center describes a pipeline that proposes candidate strategies, evaluates them against historical and market data, filters for relevance and surfaces ranked opportunities that traders accept or reject. Here is what the published material supports. Current plans are tracked on our Trade Ideas tool page.

How it works, in the vendor's words

• Overnight: Holly re-tests its strategy library on recent market data and keeps the strategies that held up. Trade Ideas' June 23, 2026 post says Holly 'runs over 60 algorithms every night.'
• During the session: it scans 8,000-plus US stocks and delivers '5 to 25 specific trade ideas each session with entry and exit levels included.' You decide whether to take them.
• Execution: automated follow-through exists only through supported broker integrations; otherwise you place the order.

The vendor's own learning-center summary flags the main risk itself: 'out-of-sample fragility and regime shifts explain why relevance filters and trader oversight remain essential.'

The reviews disagree on basic facts

The same spread shows up on signal volume: Trade Ideas says 5 to 25 ideas per session, while Alpha GainDaily counted roughly 5 to 8 signals on a normal day. The likely explanation is that the library and settings change over time and by Holly variant, which is a reason to read any review's numbers as a snapshot.

The only public test with a count

Alpha GainDaily paper-tracked Holly signals over 45 days, mid-January to late February 2026, and logged about 280 signals across the three Holly variants. It reports that roughly 62% to 64% of standard Holly signals reached target before stop, during a period with the VIX roughly 14 to 22 nContext: n=about 280 signals across 3 Holly variants, per-variant counts not stated · mid-Jan to late Feb 2026, 45 days · Alpha GainDaily paper test · margin: not disclosed. The reviewers call their own result 'directional rather than precise' because paper tracking ignores slippage, partial fills and the pressure of live trading, and they say the sample is too small for statistical certainty. They also note the company's roughly 65% accuracy claim had no independent audit as of early 2026 nContext: n: not disclosed · claim as of early 2026 · Trade Ideas, as reported by Alpha GainDaily · margin: not disclosed.

Hit rate is not profitability. A signal can reach target more often than not and still lose money once size, losers, fees and slippage are counted. We report the figure only because it is the one public, dated count we found.

What reviewers agree on

• Desktop and steep. StockBrokers.com says Holly, The Money Machine and backtesting live only in the desktop platform, which has a steep learning curve.
• Crowding. The same review warns that signals are pushed to the whole community, so popular entries can get crowded and speed matters.
• Machine learning, not a chatbot. StockBrokers.com describes Holly as practical machine learning rather than generative AI; Trade Ideas separately offers a plain-English assistant that builds scan windows.
• Simulated statistics. Several reviews stress that Holly's published statistics are simulated rather than a forward-tested public log.

What it costs (read October 14, 2026)

Source: Trade Ideas pricing. Reviews quote other figures, including $228 a month in Alpha GainDaily's February piece, another sign of how fast review prices age. Trade Ideas' pricing page calls TI Premium 'AI-driven strategies and signals to maximize profits'; its footer says it is a publisher and no content is a recommendation suitable for any specific person.

What we leave out

Some trading-tool roundups quote Holly's backtested annual returns. We do not repeat them. They are simulated, and Trade Ideas' own homepage examples, per its review page, highlight signals that 'became large long-term winners', which is a selection, not a record.

If you evaluate it

• Paper-trade a few weeks of signals yourself with realistic slippage before paying for Premium.
• Log every signal, not the ones you liked: selection is the easiest way to fool yourself.
• Check the subscription against account size; Alpha GainDaily's break-even math shows the fee alone demands a sizable return on a small account.