Search engine results giving bad info about UT
-
I learned something earlier today regarding search engine database and training (I thought it worked this way, except for a detail):
I asked Google - in Dive Deeper mode - WHY do I so often get wrong answers about things, (such as UT device functionality), in my initial query to google, and then get a completely different, & often (but not always) more accurate answer in 'dive deeper' mode.
Gemini 3.2 Pro explained that the initial response system was trained a few years ago, and not updated recently, and is intended to give fast answers.
Whereas in dive deeper mode, it actually goes out and searches fresh (or fresher ...) and gets a 'better' answer. (I sorta knew that, but I thought that newer info might be used to update 'training' for the 'quick response initial reply system. NO !)
But 'Dive Deeper' still wrong enough times to cause frustration.
One problem here is, I only knew Gemini 3.2 Dive Deeper was wrong because I recently read details on the forum, or tried something on UT phone.
G 3.2 Pro also explained that a low percentage of users dive deeper, and an even lower percentage 'push back'.
[and BTW, i 'pushed back, in dive deeper mode, a few weeks ago, on a UT device detail: lack of bands on specific phone for use in USA. It found the info, and updated its answer to me. I thought that G 3.2 Pro had 'LEARNED' new info, not from something I claimed, but from its own search. WRONG. I asked same question today, to test whether it had learned, but it gave me same wrong answer it initially did weeks ago ! BAD learning !]
So the average person looking to try Linux on their phone does a typical google search, and gets info that might raise their expectations (such as Pixel 3A is a primary supported phone for UT, & usable in USA), so they buy one (in USA), install UT, and discover no VoLTE.
So that potential UT user is now upset that they wasted time, and angry at UT.
I would say that it NOT UT fault, but a google search update fault.
But G 3.2 Pro gave me the idea that the only way the Quick Response thing will be updated is when a massive number of postings all over the web are showing updated info. And even G 3.2 Pro does not learn !
So, hopefully, the searching person lands on UT device page, and it is much more accurate, and expectations match reality, and we end up with happy new users.
I have read that people are working on updating those UT device pages.
Anybody know how to force google to update its bad info ? (I failed !) -
@oldbutndy Sorry, no idea!
I would hope that someone looking for information would get their info at the source: our webpage. Not an AI summary that may or may not be correct.
I may be oldfashioned... but I always try to look for the original source. -
AI is not a good source of information especially about niche things
-
@oldbutndy Sorry, no idea!
I would hope that someone looking for information would get their info at the source: our webpage. Not an AI summary that may or may not be correct.
I may be oldfashioned... but I always try to look for the original source.Agreed, but I restarted my search 2025 December via Google and I know I also read the UT device page (because I made a spreadsheet then, comparing the top choices, and included details such as wireless ext display).
I don't remember device page clearly saying VoLTE not working, main developer carrier NOT using VoLTE, and therefore low likelihood of P3A XL ever getting good VoLTE, and no native VOIP app for current UT release.
So, I ordered the 3A XL, installed UT (great easy process, but it IS easy on an unlocked Pixel, and I had already been through similar process with LineageOS, on P XL, and 4XL.)
I also don't remember in early December, whether Nord N10 5G was yet working for VoLTE, but by late that month I got replies to forum questions suggesting I try it (since 1 person had it working in USA on T-Mobile). That meant 'possible', at least. (so, I bought 1 !)
All I am trying to say here is we need to keep striving to make sure the main UT landing pages are accurate, especially for critical details: examples (1) do any device pages clearly state that some highly promoted devices are missing bands required by some carriers in some countries ? or (2) on another brand, if you brick the main board during UT install, you effectively cannot get a part replacement in any country but France (if I understand that thread correctly).
Those 2 facts might cut 2 of the MOST recommended options for US purchasers.
So, last I checked, Nord N10 5G is not highlighted, but if in USA, and you want a phone that works on VoLTE now, and lasts a bit into the future, Nord N10 5G is the ONLY economical viable option ?
If google quick search gets it wrong, and Dive Deeper gets it wrong, but people find UT device page and critical info NOT obvious, and new people waste time & money, isn't that at least partially our fault ?
I understand that someone IS doing their best to update the tiny details, on device pages, but I was told that is completely different from adding a sentence to a paragraph on a web page (I did not know that).
That is why I tried yesterday to see if I could improve the data that Google fed to people.
I hate feeling helpless.
I have no idea how to edit a webpage. (and I did not yet bother to ask Google about it.),
And, since I am a newbie & keep finding conflicting info all over, I would not want to edit a page unless I was POSITIVE about it. I don't want to make anything WORSE ! -
google quick search gets it wrong, and Dive Deeper gets it wrong, but people find UT device page and critical info NOT obvious, and new people waste time & money
@oldbutndy Maybe it will be possible to establish CI/CD pipeline from gitlab that will populate a "tested features" table from commints from testers forks provet and merged. Or there are even easier tools for such pipeline, as it involves no code ondy results from testers. Maybe ubports already uses something similar to it like 'ig projects have, and something got lost in the process. Dont know where to direct these qustions tho, just thinking outloud. But doing those test results tables by hand is very tedious and error prone
Hello! It looks like you're interested in this conversation, but you don't have an account yet.
Getting fed up of having to scroll through the same posts each visit? When you register for an account, you'll always come back to exactly where you were before, and choose to be notified of new replies (either via email, or push notification). You'll also be able to save bookmarks and upvote posts to show your appreciation to other community members.
With your input, this post could be even better 💗
Register Login