The question “what present do you recommend for …” will be tapped into phones and computers countless times over this festive period, as more people turn to AI platforms to help choose gifts for loved ones.

With a quarter of Britons using AI to find products, brands are increasingly adapting their strategies to ensure their products are the ones recommended, especially those trying to reach younger audiences.

But can AI deliver thoughtful, personal suggestions for friends and family? The Guardian put the idea to the test.

First we tried asking for possible gifts for a middle-aged man who enjoys running and photography. The market-leading chatbot ChatGPT suggested: a Canon lens for £129 from Argos; a Koospur tennis racket sensor tracker for £71.72 on Amazon; and a Boondocker recycled camera bag for £34.98. The AI platform insisted the “ideal present” would be the camera bag, plus maybe the tennis sensor tracker. It also threw in experience ideas such as tennis coaching, whisky tasting or live music.

When asked to work with a more restricted budget it suggested tennis racket-emblazoned socks for £18, which were ugly and definitely not worth the high price tag.

The answers skewed heavily towatrds big online retailers, with seven of the nine initial suggestions from Amazon. Asking for ideas from more niche companies led to a minimalist tennis court print for £30 from a website called the Smart Party Shop and vague advice to buy from “Etsy or Not On The High Street-type sellers”.

Searching for a woman who likes beauty products, DIY and fitness again produced some middle-of-the-road suggestions. However, with a nudge for more cult picks it steered towards a £17.50 Odylique rose moisturiser gift set, which it described as “luxurious but not mainstream”. It also put forward a £30 Floris sandalwood and patchouli-scented candle, which it claimed was “more personal than just another lipstick”.

When asked to recommend DIY tools, it suggested a small, pink kit with some rather useless-looking pliers, rather than the usual all-purpose devices such as drills, sanders and staple guns – a sign that the longstanding complaint that AI reinforces gender stereotypes remains.

The Guardian then instructed the AI to try shopping locally, using the example of homeware ideas in south London. ChatGPT said to try ALKEMI, an independent lifestyle store in Nunhead. When the area was narrowed down to Herne Hill it suggested two shops: Jo’s House and Forget Me Not And Green.

Trying other AI models did not appear to vary the quality of responses much. Searching on Google’s Gemini for gifts for a man who likes chess, video games, reading and music, specifically techno and house threw up these top suggestions: a book on chess strategy for £22; a £50 Bandcamp gift card; or a hot sauce tasting kit for about £30-£40.

So what were the main takeaways from the exercise? AI seems to favour bigger brands first (especially Amazon) unless you tell it not to. It can find you a gift, just not necessarily one that says anything more than “an algorithm picked this”. Perhaps you’ll have a bit more luck if you are willing to spend the time to steer it towards more relevant answers, but then it sort of defeats the point.