Why asking ChatGPT what to buy isn't the same as getting a real recommendation
If you've ever typed "what's a good alternative to Whispering Angel?" into ChatGPT, you'll know the feeling — a confident, well-structured answer that somehow tells you very little. The AI product recommendation experience has a problem: most AI tools are built to answer questions, not to find you the right bottle of rosé from a shelf you can actually buy from, today, in the UK. That's exactly the gap More Like This was built to close.
This post is about how our AI actually works, what it's been shaped to do differently, and why the distinction matters when you're standing in front of a screen trying to decide what to spend your money on.
What More Like This is actually built on
More Like This is powered by Anthropic's Claude — one of the most capable large language models available right now. Claude was chosen specifically because of how it handles nuance, context, and honest reasoning. It doesn't hallucinate product names with the same enthusiasm as some alternatives. It's cautious with confidence. And it's particularly good at understanding what someone means, not just what they literally typed.
When you search for "something like Glenfiddich 12 but a bit more interesting", Claude doesn't just pattern-match on "Scotch whisky, 12 year old, affordable". It reads intent. It understands that "a bit more interesting" is doing real work in that sentence — and it looks for something that genuinely moves the needle, rather than serving you a list of similar-sounding bottles.
But the model is only part of it
A general-purpose AI — even a very good one — doesn't automatically become a great product recommendation engine. The reason More Like This works differently from dropping a question into ChatGPT comes down to three things: focus, context, and live inventory.
Focus: More Like This has been configured specifically around UK product discovery. It knows The Whisky Exchange's range. It knows the difference between what Majestic stocks and what Master of Malt stocks. It understands UK pricing norms, UK seasons, UK gifting culture. It's not trying to also help you write a cover letter or explain quantum physics. It does one thing.
Context: The platform feeds Claude structured, category-specific context before it ever generates a recommendation. That means it's not working from general internet knowledge alone — it's working with real, up-to-date understanding of what's available, what's well-regarded, and what's genuinely worth your attention in each category.
Live UK retailer search: This is the part that changes everything. ChatGPT, in its standard form, doesn't search Selfridges or Fortnum & Mason in real time. It works from training data with a knowledge cut-off. More Like This actually searches live UK retailer inventory — so when it recommends a bottle of Nikka from the Barrel or a Mirabeau Provence Rosé, it's checking that the product exists, is available, and can be found at a UK retailer right now. Not "probably available somewhere". Actually there.
The three-result structure — and why it's deliberate
Most recommendation tools give you a list. More Like This gives you three specific results, each with a distinct purpose:
- Closest Match — the recommendation that most faithfully mirrors what you searched for. If you want something genuinely similar, this is it.
- Second Match — a well-matched alternative that approaches the same appeal from a slightly different angle. Same territory, different path.
- Wildcard — something you probably wouldn't have found yourself. Not random — deliberately chosen because it shares the same underlying appeal, just packaged differently.
This structure isn't arbitrary. It reflects how good recommendations actually work in real life. A knowledgeable friend doesn't hand you a spreadsheet — they say "okay, if you want something really close, try this. But if you're feeling adventurous, you should know about this other one." The Wildcard in particular is where More Like This tends to earn its keep. It's the Teeling Small Batch when someone searches for Jameson's. The Château Minuty when someone's always bought the same supermarket rosé.
What it doesn't do — and why that honesty matters
More Like This isn't magic, and it's worth being straight about that. It doesn't have opinions shaped by years of actually tasting whisky or wearing a fragrance. It can't tell you whether a candle smells better cold or lit. What it can do is synthesise an enormous amount of real product knowledge, reviewer consensus, flavour profile data, and retailer information — and use that to make a genuinely useful call about what you'd probably like.
There are no paid placements in the results. No affiliate arrangements that push certain products up the list. The AI doesn't know which bottle has the best margin for a retailer — and it wouldn't care if it did. The recommendation is made on fit, not on commercial interest.
Why this beats a generic ChatGPT prompt
To put it plainly: ChatGPT is a generalist. An exceptionally capable generalist, but a generalist nonetheless. If you ask it for a whisky recommendation, it'll give you something reasonable — probably Auchentoshan, Oban, something from Bruichladdich. But it won't check whether that bottle is £28 or £280 at a UK retailer right now. It won't frame the answer around three distinct recommendation angles. And it won't have been specifically shaped around the nuances of UK drinks culture, UK skincare retail, or what's actually on the shelf at Harvey Nichols this week.
More Like This is a specialist. And in most areas of life, when you need a real answer, you want a specialist.
Try it yourself — it takes about thirty seconds
The best way to understand how the AI works is to use it. Type in something you already love — a whisky, a perfume, a skincare product, a candle — and see what comes back. Not because the results will always be perfect (no recommendation engine is), but because the approach is genuinely different from anything else currently available to UK shoppers.
If you've been relying on Google searches, supermarket algorithms, or the occasional ChatGPT prompt to find new products, try More Like This and see what a purpose-built UK discovery tool actually feels like.
It's free. It's fast. And it searches real shelves.