I have been experimenting with Chat GPT and uploaded a spreadsheet the wines I hold in bond and asked it which wines I should prioritise for drinking.
It proposed that the wines that had been in bond longest should be drunk first.
I wonder, having very limited technical ability, if it is possible to refer it to a ‘reputable tasting database’?
1 Like
Have you tried to be more didactic or detailed in your prompts? For example, rather than saying “which wines should i drink first”, try asking something more along the lines of “Which of these wines should I prioritise drinking soonest, based on recommended drinking dates for the given vintages, as published by sources including but not limited to…”
2 Likes
Yes, that’s probably key. It can be quite helpful (or quite unhelpful) depending on context and the prompt used. If you have a PDF or other file with a list of wines and drinking dates, you can upload that and get it to use it – but even without that, a better prompt might give you better results.
I’ve asked it for advice on pairing wine with meals (or vice versa) and it’s often quite good at that.
I tried copying my CT list across with this prompt:
Can you please give me proposed drinking dates for the wines in my cellar? (below)
Thanks!
It’s a bit simplistic but kinda works. It does take the wine style into account to some extent. I then followed that up with:
Can you give specific recommendations based on the specific wines? Thanks
This gave a bit more detail, but some suggestions seemed too early for me still.
Also, I know it’s odd but please and thankyou really help improve its output. It responds in a kinda human like way, and polite, helpful questions trigger the context for polite, helpful responses.
Also, telling ChatGPT “If you don’t know, don’t guess” can sometimes have an enlightening impact on results!
3 Likes
Thanks for the suggestions. In my experience Chat GPT is an amazing resource though capable of giving an incorrect answer. I will continue to experiment.
1 Like
My experience is similar. It’s an amazing resource, but there are some things it excels at and others it’s not so great at.
As an example of the latter, it was reported recently that ChaGPT 4o was pitted against a 1977 Atari 2600 (1.19Mhz!) running the 1979 Atari Chess game, and lost. With the 4o model, at least, that kind of reasoning over a longer period (multiple prompts) is not something it’s great at.
And yes, if using for anything important, it’s best to double-check its suggestions as it can make mistakes.
Try using VitaBella. It’s chat GPT specifically for wine. You can find it in the explore GPTs section of the app. Let us know if the results are any different
1 Like
Sorry, stumbled on this thread and thought I’d share my experience of ChatGPT 5.1 and my wine cellar.
My prompt was: Which of these wines should I prioritise drinking soonest, based on recommended drinking dates for the given vintages, as published by sources including but not limited to Neal Young, Decanter, Jancis Robinson and others.
The summarised results out of 160 wines were:
| Wine | Vintage | Category | Drink Window | Priority |
|---|---|---|---|---|
| Châteauneuf-du-Pape Blanc ‘Le Miocène’ (Alain Jaume) | 2016 | White | 2017–2020 | Drink Now |
| Bourgogne Blanc Cuvée Oligocène (Patrick Javillier) | 2017 | White | 2020–2022 | . Drink Now |
| Old Vine White (Mullineux) | 2015 | White | 2020–2024 | Drink Now |
| Viña Bosconia Reserva (López de Heredia) | 2005 | Red | 2018–2025 | Drink Now |
| Skurfberg (Sadie Family) | 2016 | White | 2018–2028 | Drink Soon |
| Reserve Chenin Blanc (DeMorgenzon) | 2017 | White | 2020–2028 | Drink Soon |
| Rocking Horse Cape White (Thorne & Daughters) | 2017 | White | 2021–2028 | . Drink Soon |
| Clos des Papes Châteauneuf-du-Pape | 2012 | Red | 2015–2028 | Drink Soon |
| Viña Ardanza Reserva (La Rioja Alta) | 2012 | Red | 2021–2030 | Drink Soon |
| Château Batailley, Pauillac | 2009 | Red | 2018–2035 | Hold |
| Château Batailley, Pauillac | 2010 | Red | 2020–2040 | Hold |
| Château Le Crock, St-Estèphe | 2010 | Red | 2018–2038 | Hold |
| Château Gloria, St-Julien | 2010 | Red | 2018–2038 | Hold |
| Rubicon, Meerlust | 2009 | Red | 2020–2032 | Hold |
| Sijnn Red Blend, Malgas | 2010 | Red | 2018–2035 | Hold |
| Viña Tondonia Reserva (López de Heredia) | 2006 | Red | 2017–2031 | Hold |
| Viña Tondonia Reserva (López de Heredia) | 2011 | Red | 2023–2039 | Hold |
I can’t disagree with any, bar maybe the Sadie Skurfberg - 2016 feels still a bit young.
Overall I was pretty impressed.
The estimate for the drinking window of the Viña Bosconia is likely to be too conservative in my experience. I recently had a 2007 and a 1999 of this wine and they were both drinking beautifully. They did not yet seem to be approaching the end of their drinking window.
1 Like
Suspect you may mean my friend Neal Martin. I don’t think singer Neil Young claims any relationship to wine expertise, although AI search suggests otherwise:
the famous singer-songwriter
Neil Young and his half-sister Astrid Young , a musician and certified sommelier (wine expert). The search results also mention a song on which Neil Young used a wine glass as an instrument.
2 Likes
Well, thanks for correcting me, Jancis. I have to admit, I’m a fan of both Neil’s! 
1 Like
Thank you - I will resist the temptation to grab one over Christmas then.
I will try a bottle of Clos des Papes 2012 but I can’t believe that it should be drunk soon