Totally valid response :-) And I think that's another great angle as well that you can bring to a conundrum like Gmail or Google Search, like Time to Value or Engagement, etc.
As an extension, what we're seeing with OpenAi et al. is that they are capturing that attention and taking search with them. And so (as I referred to a couple comments down) OpenAI and the others are in the Google pre-2006 moment where the products are highly successfully engaging and grabbing our attention, but they haven't quite found the business model that prints money in the way Google Ads do.
This has got to be one of the most loaded question I have ever seen on here. Could you perhaps be clearer with what you want to ask? It sounds like you are insinuating that the guy is a bad person in some vague, nondescript way.
Agreed. And I added an update to give a more thoughtful response. I've been on the receiving side of trick questions, so I do understand. How you ask the question, which I didn't phrase here, is important. Google Search or Gmail are great case study style questions. Here's an oldie but a goodie about it from 2006:
Yup, but my earlier comment was a bit flippant, so here’s a more in-depth response.
It was an answer to the OP’s question about why Gmail search feels broken, and a nod to the previous comment pointing out Google’s core business is selling ads. In that context, Gmail isn’t optimized for superior search. It’s part of a data funnel that enables keyword targeting, ad placement, and behavioral profiling.
For me, there’s a deeper product lesson here, but also a signal about how someone thinks. Whether a candidate answers “ads” or “search” or “email” isn’t what really matters. It’s the why behind the answer that matters most.
At the end of the day, business models directly shape product decisions. That tension is something every product manager has to navigate because they sit between the business and technical sides of a company. Being able to see and articulate that tradeoff, especially when user needs diverge from monetization incentives, is key to both building effectively and being a successful product manager.
More often the business model, like subscriptions, is more tightly connected to the product. User has X problem that product Y solves and the company receives Z dollars in exchange for access. However, there are many examples where the pricing or monetization strategy is not so clearly connected to the feature set, of which google search is a great example.
For the record, I don't ask this question to junior PMs.
You are doing your job well, and for that I salute you.
However, you highlight the real problem with ad-supported tech. It creates perverse incentives that makes the world an objectively worse place for most just to sell another ad. It justifies actions and data collection that would be illegal if it were anything other than a large corporation peforming that activity. At some point in history the words, "just trying to make my quarterly numbers", will be looked at with the same level of horror and shame as a similar phrase was over 75 years ago.
I'm assuming whatever they're currently doing involves machine learning so uses more compute than the simple correct solution that a CS freshman could code up.
I've been on email since 1985 and had the same personal address since 1991. I've used about 20 or 25 email clients across a dozen OSes (counting all Linux distros as one).
Gmail has the best search of any email system I've found.
I think your problems are nothing to do with Gmail or its search, but are to do with things like Unicode character encoding, character sets and codepoint matching.
I once was looking for a video I watched a few months ago with a history search. Nothing brought it up.
I found it with text match in my browser history.
Out of curiosity, I tried every combination of words in its title, including the full title verbatim, and it did not show up. It is truly astonishing how bad it is.
While I was debugging this exact pain-point I prototyped a tiny web viewer that visualizes Gmail search results like a diff, highlighting every literal match. I ended up turning it into a weekend side-project powered by the same Black Forest Labs tech behind FLUX Kontext (https://flux-kontext.io/)
Typically will use Gmail in the browser for sending and recieving emails…but for searching my Gmail, it’s Outlook, ironically, which works the best.
Don’t trust Outlook for other functions in Gmail reliably though. Or Calendar. Examples are things like double emails being sent and calendar updates being missed.
At our company a lot of people use Outlook to handle their Gmail functions and it’s quite problematic…but Outlook’s search is far superior.
So I think of the browser as the main interface and only use Outlook for search.
> but for searching my Gmail, it’s Outlook, ironically, which works the best.
How bad I find the search in Outlook and Teams is one of the reasons I refuse to trust Bing. I know they are completely separate divisions, but still…
Both they and gmail have trouble finding someple words and sentence fragments that I know are in there and, worse, instead of just saying they can't find it they often return a bunch of useless stuff just-in-case I meant that instead.
I keep thinking of writing an IMAP->SQL interface¹ so I can search more manually that way. Or even just mapping to files in a filesystem² so I can find/grep/etc, though the DB option would also allow some more complex searches (search for X in messages that are replies to messages containing Y, for example).
--------
[1] Keeping mail on the mail server, but syncing it with a DB. This is more complicated than it would first sound as it will mean decoding MIME and various sub-encodings to extract the plain text so I can properly query it, hence that project is still on my “when hell freezes over so I have enough free time” list.
[2] Yeah, maildir and similar storage formats already do this, but there is still the unpacking/decoding issue before being able to directly search the text of many messages.
Gmail's search likely tokenizes/normalizes Unicode symbols like ₹ differently than plain ASCII, so try searching for "rupee" or use mail filters with regex capabilities instead.
That seemed plausible so I tried it and after more than 10 seconds, a search for "rupee" returned similarly poor results.
This is really only the latest example of nearly useless search results.
I genuinely don't understand how in terms of both latency and accuracy, Google is failing at this embarrassingly parallel problem. Fuzzy searching I'd understand, but not searches for specific strings.
This is why it’s worth paying for email not at Google. It’s not expensive.
You could self host but that’s a nuisance I’m happy to pay someone for.
Search sucks even when using English. It fails to find emails I know have certain words in the subject.
The entire Google ecosystem is a hot dumpster fire of garbage that doesn’t help me at all at this point. It used to be amazing when they focused on organizing information rather than selling eyeballs. But all things turn to shit chasing profits.
Sorry to read that you’re dealing with this. My primary email is apart of some vast international spam program where the mail servers responsible for sending me spam forge headers, falsify source record information, lie about which downstream service provider is hosting them, etc. It’s really quite interesting, but more so than that, it’s annoying.
I pondered my problem for a few days, thinking about what sort of external service I could use to surgically remove these numerous, daily, very specifically identifiable spam emails before I stumbled on a thread on StackOverflow where some people discussed using Google App Scripts to do this very thing.
I’d recommend searching the web for that sort of topic. You’ll find that there’s a way to set up an hourly script job that will wipe this spam completely off your mailbox and find some peace.
Hope this points you in the right direction. The idea of having to use yet another Google service to fix an existing separate one is such a stupid labyrinthine experience, but at least it beats having to set up a job on a VPS for this.
If you had this issue with another mailbox service provider, a VPS approach would probably be necessary, though.
Yep, it happened to Google Search first and slowly creeped to other products now. It's not even "AI", it's just incompetence.
Try this now, go to your inbox, filter emails by date, you will get back a list where your emails are sorted randomly ... no really, check it out. This is a new "feature". Their PMs should be shot.
Gmail is a toy, not a real email user agent.¹ You might be happier with Mutt, which is powerful, flexible, and fast. But probably the most efficient way to accomplish what you want is to use a tool like grepmail (assuming Google provides a way to export your mail into an mbox file or something).
idk… I for one like the fact that Gmail sucks at search to be honest because I (naively?) believe they don’t profile everything in your inbox. For example, if I subscribe to a newsletter about “healthy lifestyle,” it won’t return that newsletter, but return string matches where “healthy” or “home” are relevant. If they profiled the emails for contextual awareness to know what I meant by “healthy living,” I’d be concerned.
It seemingly does. It's part of their enterprise offering, which they can use to upsell other things like GDocs and company GDrive. GDocs is also notoriously bad with search. As bad as it is, it's still the best out there that most people are familiar with.
Yeah, I switched to DuckDuckGo years ago (when it came out that Google was "customizing" search results per-customer, meaning you'd not just get different advertising but different *search results* based on what ads you'd previously clicked on and/or what other things you'd searched for) and never looked back.
> Why has a search company
Google was a search company, many years ago.
Today's Google is an advertising company that just happens to have a legacy search division.
This. As an interview question for my product managers, I often ask what Google search's product is. The ones who say ads move to the next round.
The product of Search is data and attention. They use this data and attention to sell their Ads product.
The same can be said for almost everything they make.
If Google's customers are the advertisers, maybe the correct answer is "attention".
Totally valid response :-) And I think that's another great angle as well that you can bring to a conundrum like Gmail or Google Search, like Time to Value or Engagement, etc.
As an extension, what we're seeing with OpenAi et al. is that they are capturing that attention and taking search with them. And so (as I referred to a couple comments down) OpenAI and the others are in the Google pre-2006 moment where the products are highly successfully engaging and grabbing our attention, but they haven't quite found the business model that prints money in the way Google Ads do.
So we'll see. What do you think?
Are you aware what you are saying about yourself by writing such a sentence?
This has got to be one of the most loaded question I have ever seen on here. Could you perhaps be clearer with what you want to ask? It sounds like you are insinuating that the guy is a bad person in some vague, nondescript way.
Agreed. And I added an update to give a more thoughtful response. I've been on the receiving side of trick questions, so I do understand. How you ask the question, which I didn't phrase here, is important. Google Search or Gmail are great case study style questions. Here's an oldie but a goodie about it from 2006:
https://www.hbs.edu/faculty/Pages/item.aspx?num=33707
Trick questions like those have a terrible signal to noise ratio, you'd hope people with recruiting decision power would have gotten the memo by now.
Yup, but my earlier comment was a bit flippant, so here’s a more in-depth response.
It was an answer to the OP’s question about why Gmail search feels broken, and a nod to the previous comment pointing out Google’s core business is selling ads. In that context, Gmail isn’t optimized for superior search. It’s part of a data funnel that enables keyword targeting, ad placement, and behavioral profiling.
For me, there’s a deeper product lesson here, but also a signal about how someone thinks. Whether a candidate answers “ads” or “search” or “email” isn’t what really matters. It’s the why behind the answer that matters most.
At the end of the day, business models directly shape product decisions. That tension is something every product manager has to navigate because they sit between the business and technical sides of a company. Being able to see and articulate that tradeoff, especially when user needs diverge from monetization incentives, is key to both building effectively and being a successful product manager.
More often the business model, like subscriptions, is more tightly connected to the product. User has X problem that product Y solves and the company receives Z dollars in exchange for access. However, there are many examples where the pricing or monetization strategy is not so clearly connected to the feature set, of which google search is a great example.
For the record, I don't ask this question to junior PMs.
You are doing your job well, and for that I salute you.
However, you highlight the real problem with ad-supported tech. It creates perverse incentives that makes the world an objectively worse place for most just to sell another ad. It justifies actions and data collection that would be illegal if it were anything other than a large corporation peforming that activity. At some point in history the words, "just trying to make my quarterly numbers", will be looked at with the same level of horror and shame as a similar phrase was over 75 years ago.
How does compromising the ability to search your inbox increase ad revenue?
Probably made other tweaks that gave marginal gains in ad revenue whilst neither caring nor measuring about the effect on search results.
Presumably ad revenue has gone up even as search result quality has gone down, proving that the ability to search your inbox is actually unimportant.
Computation costs probably not worth it for the incremental user satisfaction.
I'm assuming whatever they're currently doing involves machine learning so uses more compute than the simple correct solution that a CS freshman could code up.
Is probably less about making as revenue and more about not investing in it
gmail search was never good though.
I've been on email since 1985 and had the same personal address since 1991. I've used about 20 or 25 email clients across a dozen OSes (counting all Linux distros as one).
Gmail has the best search of any email system I've found.
I think your problems are nothing to do with Gmail or its search, but are to do with things like Unicode character encoding, character sets and codepoint matching.
I find Gmail search to be OK. Compared to search functions in other tools it’s probably near the top. I’m looking at you Notion…
Curious if the poor search performance you saw is related to the non-Roman alphabet search or another factor.
I've found Gmail search to be suprisingly good. It can parse pdf attachments from years ago and match it with the search.
It's miles ahead of Outlook from my point of view.
It's still better than YouTube's search.
I once was looking for a video I watched a few months ago with a history search. Nothing brought it up.
I found it with text match in my browser history.
Out of curiosity, I tried every combination of words in its title, including the full title verbatim, and it did not show up. It is truly astonishing how bad it is.
Ugh that's a whole other rant.
If I know what channel the video I'm looking for is from, I end up going directly to the channel and searching from there.
Otherwise I've definitely resorted to using my history as well.
While I was debugging this exact pain-point I prototyped a tiny web viewer that visualizes Gmail search results like a diff, highlighting every literal match. I ended up turning it into a weekend side-project powered by the same Black Forest Labs tech behind FLUX Kontext (https://flux-kontext.io/)
Is this an ad?
Typically will use Gmail in the browser for sending and recieving emails…but for searching my Gmail, it’s Outlook, ironically, which works the best.
Don’t trust Outlook for other functions in Gmail reliably though. Or Calendar. Examples are things like double emails being sent and calendar updates being missed.
At our company a lot of people use Outlook to handle their Gmail functions and it’s quite problematic…but Outlook’s search is far superior.
So I think of the browser as the main interface and only use Outlook for search.
> but for searching my Gmail, it’s Outlook, ironically, which works the best.
How bad I find the search in Outlook and Teams is one of the reasons I refuse to trust Bing. I know they are completely separate divisions, but still…
Both they and gmail have trouble finding someple words and sentence fragments that I know are in there and, worse, instead of just saying they can't find it they often return a bunch of useless stuff just-in-case I meant that instead.
I keep thinking of writing an IMAP->SQL interface¹ so I can search more manually that way. Or even just mapping to files in a filesystem² so I can find/grep/etc, though the DB option would also allow some more complex searches (search for X in messages that are replies to messages containing Y, for example).
--------
[1] Keeping mail on the mail server, but syncing it with a DB. This is more complicated than it would first sound as it will mean decoding MIME and various sub-encodings to extract the plain text so I can properly query it, hence that project is still on my “when hell freezes over so I have enough free time” list.
[2] Yeah, maildir and similar storage formats already do this, but there is still the unpacking/decoding issue before being able to directly search the text of many messages.
I was asking myself the same question the other day, some basic search thing just didn't work.
Finding things in Google Drive is an awful experience as well.
Gmail's search likely tokenizes/normalizes Unicode symbols like ₹ differently than plain ASCII, so try searching for "rupee" or use mail filters with regex capabilities instead.
That seemed plausible so I tried it and after more than 10 seconds, a search for "rupee" returned similarly poor results.
This is really only the latest example of nearly useless search results.
I genuinely don't understand how in terms of both latency and accuracy, Google is failing at this embarrassingly parallel problem. Fuzzy searching I'd understand, but not searches for specific strings.
In 2025!
If you think that's bad wait til you try Outlook.
This is why it’s worth paying for email not at Google. It’s not expensive.
You could self host but that’s a nuisance I’m happy to pay someone for.
Search sucks even when using English. It fails to find emails I know have certain words in the subject.
The entire Google ecosystem is a hot dumpster fire of garbage that doesn’t help me at all at this point. It used to be amazing when they focused on organizing information rather than selling eyeballs. But all things turn to shit chasing profits.
Sorry to read that you’re dealing with this. My primary email is apart of some vast international spam program where the mail servers responsible for sending me spam forge headers, falsify source record information, lie about which downstream service provider is hosting them, etc. It’s really quite interesting, but more so than that, it’s annoying.
I pondered my problem for a few days, thinking about what sort of external service I could use to surgically remove these numerous, daily, very specifically identifiable spam emails before I stumbled on a thread on StackOverflow where some people discussed using Google App Scripts to do this very thing.
I’d recommend searching the web for that sort of topic. You’ll find that there’s a way to set up an hourly script job that will wipe this spam completely off your mailbox and find some peace.
Hope this points you in the right direction. The idea of having to use yet another Google service to fix an existing separate one is such a stupid labyrinthine experience, but at least it beats having to set up a job on a VPS for this.
If you had this issue with another mailbox service provider, a VPS approach would probably be necessary, though.
Yep, it happened to Google Search first and slowly creeped to other products now. It's not even "AI", it's just incompetence.
Try this now, go to your inbox, filter emails by date, you will get back a list where your emails are sorted randomly ... no really, check it out. This is a new "feature". Their PMs should be shot.
Gmail is a toy, not a real email user agent.¹ You might be happier with Mutt, which is powerful, flexible, and fast. But probably the most efficient way to accomplish what you want is to use a tool like grepmail (assuming Google provides a way to export your mail into an mbox file or something).
[1]https://lwn.net/Articles/837960/
idk… I for one like the fact that Gmail sucks at search to be honest because I (naively?) believe they don’t profile everything in your inbox. For example, if I subscribe to a newsletter about “healthy lifestyle,” it won’t return that newsletter, but return string matches where “healthy” or “home” are relevant. If they profiled the emails for contextual awareness to know what I meant by “healthy living,” I’d be concerned.
my answer is a question: does email make a lot of money for google?
It seemingly does. It's part of their enterprise offering, which they can use to upsell other things like GDocs and company GDrive. GDocs is also notoriously bad with search. As bad as it is, it's still the best out there that most people are familiar with.
It likely doesn't make any money by itself. It's a loss leader to entice companies into its enterprise offerings.
Loss leaders are valuable too. Android is a loss leader that funnels half the world into their exorbitant Google Pay + Play Store ecosystem.
Because Google is crap at basic search these days.
Yeah, I switched to DuckDuckGo years ago (when it came out that Google was "customizing" search results per-customer, meaning you'd not just get different advertising but different *search results* based on what ads you'd previously clicked on and/or what other things you'd searched for) and never looked back.