Is Google really that stupid?

Martin Larsen

Active Member
Licensed User
Longtime User
I have made an app for the Danish Cat Protection Society that is currently only in internal testing.

The app is used for:

- Keeping a record of your cats including age, picture, chip number and the vet's phone number etc
- Finding the nearest vet
- Adoption of cats
- Information on how to deal with emergencies such as finding a wounded cat
- Showing news and magazines

Then a week ago I got an email from Google saying that the app was suspended and removed due to violating their policies. Especially:
About the Sensitive Events policy
We don't allow apps that lack reasonable sensitivity towards or capitalize on a natural disaster, atrocity, conflict, death, or other tragic event.


I used their appeal form and kindly explained what the app did and that I believed that it was flagged because of a mistake. Because there was no references to aforementioned kind of events.

Yesterday I got a reply from Jessica:
For example, your app currently contains references to the recent sensitive event (e.g. COVID-19). Please review the entirety of your app to remove all mentions to COVID-19 synonyms (including in app, all text in the description and text within screenshots).


So it was because of referring to COVID-19! But here comes the extremely stupid part:

The only reference to COVID-19 is in a webview showing news and practical information from the website. It only tells that the office is closed due to the pandemic and that the planned general meeting is canceled for the same reason.

Here is a translated version of the news page.

So I wrote back to Jessica with two questions:
  • Is such use really a problem? We would think that it is important for the members to know that the office is closed and that there is no general meeting.
  • If it is indeed a problem, would it be OK to link to the news page externally instead of opening it in a webview?

What I got back today was a (probably) canned response saying:

Unfortunately I'm not able to comment on your planned implementation. If you think your app is in compliance, please make appropriate changes to your app and submit it for another review.

Kindly note we are currently only approving apps that reference COVID-19 or related terms in their store listing if the app is published, commissioned, or authorized by an official government entity or public health organization. You may want to review the entirety of your app to remove all mentions to COVID-19 synonyms (including in app, all text in the description and text within screenshots).


So much for Google's AI. It is clearly not AI flagging the app, it is simple keyword searches.

Have you experienced the same kind of silliness from Google?

And what to do? Do you think it would be ok just to open the news page externally in a browser or would that also count as mentioning COVID-19?
 

Sandman

Expert
Licensed User
Is Google really that stupid?
Yes. Depending on how you define "stupid", of course. Regardless of AI or keyword searches, I think it's useful to think of the Play Store(*) as something that is 99,999999% automated. On top of that they sprinkle a handful of humans to handle appeals. These humans operate within a set of very narrow restrictions, and when something is outside those restrictions they can neither comment on anything nor actually help you.

And what to do? Do you think it would be ok just to open the news page externally in a browser or would that also count as mentioning COVID-19?
If you open the page externally they might consider it to be the same as you are doing now (I'm sure other people here in the forum will know more about that solution).

The least-effort solution that I can think of is to keep the webpage internal as it is now, but slap something extra onto the url. So if you previously had the url https://www.example.com/index.php, now you would instead show https://www.example.com/index.php?origin=app. (Because the app user doesn't see the url anyway, right?) And on your website you would hide the COVID-19 info when you get that parameter. Result is that all app users (and Google's tests) wouldn't see that info, but ordinary visitors to your site would see it.

(*) And almost all other Google Services. Except Google Ads, of course. That's the only place I know where they have any real amount of staff available to help the advertisers spend more money with them. I use them, and my contact person is always oh-so-happy to help me with my ads and increase my spending.
 

Martin Larsen

Active Member
Licensed User
Longtime User
The least-effort solution that I can think of is to keep the webpage internal as it is now, but slap something extra onto the url

Yes, that is a solution I have considered myself. And I wouldn't even have to add something to the url because the app uses a custom user agent already. We use it for presenting the various pages in a simpler way. For example, we remove the headers and footers because they are redundant when opened in the app and only bring clutter.

Technically it is not too difficult but I find it sad that we can't even tell the users that the office is closed! But it is probably the best thing to do right now.

Because the app user doesn't see the url anyway, right?

Correct.
 

Alexander Stolte

Expert
Licensed User
your app will be blocked while other apps, including a google app:
facebook:
Screenshot_20200422-114202.jpg
youtube:
Screenshot_20200422-114140.jpg
tiktok:
Screenshot_20200422-114015.jpg

I find it rather a pity that the app is directly blocked without having the chance to fix the bug
 

Sandman

Expert
Licensed User
we can't even tell the users that the office is closed
Well, you could probably say something like "For the time being our office is closed". Remember what the bot/humanoid said: You may not refer to COVID-19. But I doubt it can be a crime to say that your office is closed.
 

Martin Larsen

Active Member
Licensed User
Longtime User
I have found that it is bit more difficult than expected because we don't know how Google access the app. We can control the pages that are loaded directly in the webview, but what if Google follow links to other pages within the webview? Then they can and will eventually land on one of the pages that mention COVID-19 or coronavirus.

We already hide the headers and footers as mentioned before because it gives a cleaner look. But we hide them with CSS. That means that eg. the main menu has an invisible menu item with COVID-19 information. The user can never see or click on that menu item. But if Google instead just look at the html content instead of what is actually visible, that menu item may be a problem.

Then we have a lexicon with various cat diseases. One of these is Infectious Peritonitis which is caused by a coronavirus! Yes, a coronavirus, but not SARS-CoV-2. So do we really have to remove that page from our lexicon in order not to offend Google and get kicked out again?

Oh my god they are stupid.
 

rabbitBUSH

Well-Known Member
Licensed User
I suppose @Alexander Stolte [ #4 ] has the answer : they can't read an image can they? so if every tim eyou want to mention The Virus just place an image of the word, , , ,
May be more than one way to skin a Google.
Oh my god they are stupid.
Aren't all corporate sell outs. . . .[no ? no question no new thread . . . skinned]
 

Martin Larsen

Active Member
Licensed User
Longtime User
It's not that simple. They explicitly ask us to remove all mentions of the dirty words in all texts, meta data, images and so forth.

Also it's not simple technically. The backend is a Drupal driven website and we can't just replace all the occurrences with images. It would take a lot of work.

If I knew it was only the single page we load directly in the webview, it would be easy. But we don't know if Google crawls the pages and eventually lands on the "infected" pages.
 

rabbitBUSH

Well-Known Member
Licensed User
But we don't know if Google crawls the pages
look at the robot stats for the site you'll probably find that they've gone behind the firewalls and snuffled the content ... irritates me when my sysadmin wants a permissions lockdown to stop invasions but can't answer the question how come the robots can do it?
 
Last edited:

Martin Larsen

Active Member
Licensed User
Longtime User
Yes the case for COVID-19 is described in that link, but even if I had read it, I would never for a second have thought that merely mentioning that the office is closed due to COVID-19 would be a problem. And then there is the case with the cat disease caused by a coronavirus...

The reason I call Google stupid is not that they have these rules as they are both sane and needed. But they are stupid because their $$$$$ super duper AI cannot tell that we just say the office is closed due to the pandemic. Frankly, I had higher thoughts about them.
 

rabbitBUSH

Well-Known Member
Licensed User
access to the logs but I will ask the client.
you don't need access to logs - usually the client's ISP / hosting will provide a stats page like AWStats or something as a graphic and table based review so that the client can tweek and tailor their sites and content. which obviously is based on the logs for the client's account only. if the ISP doesn't provide this then the client can usually attach a visit tracker to their own pages. [[ i used to use one like this : revolvermaps 3D widget. ]] buuuut you already might know all that .........
the cat disease caused by a coronavirus..
what happens if one formats the words Corona and Virus and coronavirus -> C o r o n a V i r u s -> c o r o n a v i r u s? Just to mess with their word checker.

somehow I doubt its anything to do with Ai. Google started out by being good at word search and scanning a bazillion web pages and sites, so one might think that their system just looks in the APK for words and flags them regardless of the semantics of the sentence. the same as one would four letter words (or whatever length these might be in whatever language).

the more alarming part is that your appeal fell out of the hands of the warm-blood component so quickly. this is typical of call centers or help lines where there are many calls and requests. seems like larry and sergei have lost their way there.
 
Last edited:

JohnC

Expert
Licensed User
Longtime User
What about downloading the pages in webview or okhttputils then doing a .Replace("covid-19","current event") before showing them in webview.

This way any webpage shown in your app (using webview) will not have that "hot" word in it (because it is substituted with plain words) so they wont be flagged.
 

udg

Expert
Licensed User
Longtime User
I would do the opposite. On the original website, write Covid-19 as two side by side images; first one showing "Cov" while teh other shows "id-19".
Their Ai will evaluate each image per se missing the combined meaning..probably.
 

Martin Larsen

Active Member
Licensed User
Longtime User
what happens if one formats the words Corona and Virus and coronavirus -> C o r o n a V i r u s -> c o r o n a v i r u s?

That would most likely not matter. Google is very good at finding pages based on wrong and even weird spellings. Furthermore, it may backfire if Google deems the effort deceptive. This applies to using two images side by side etc. Also, it takes many resources for so little.

the more alarming part is that your appeal fell out of the hands of the warm-blood component so quickly

Very true.

What about downloading the pages in webview or okhttputils then doing a .Replace("covid-19","current event") before showing them in webview.

Yes, that is something I have considered too. The advantage is that we don't have to modify the CMS. And it is not deceptive.
 

rabbitBUSH

Well-Known Member
Licensed User
What about downloading the pages
But.... I wonder if all this is happening before the app gets onto the platform and can even get to download the offending pages?
Their Ai will evaluate each image per se missing the combined meaning..probably.
Here's the thing.... To evaluate those images and do the ai would mean a lot image processing and serious algorithms to figure out if an image contained certain designated words. While they're evidently and eminently capable of doing that, it's doubtful they do more than just a word search, just in a volume. It seems the rules leave it to the developer to manage a bunch of this stuff if you read earlier messages from @Martin Larsen here. That's just a warning to say 'if we catch you thennnnn...'.
 

Martin Larsen

Active Member
Licensed User
Longtime User
But.... I wonder if all this is happening before the app gets onto the platform and can even get to download the offending pages?

Good question. There are two possible ways Google could evaluate an app. One is to analyze the apk and in this case look at which urls are loaded into the webview. The other way is to run the app in a emulator and crawl all buttons and menu actions etc. I think it is the latter.

It might be possible to tell by looking at the logs. I prefix the normal user agent in order to serve the content in a special way. And since only 5 or 6 devices has ever installed the app so far (it's in internal testing, we should be able to tell from the logs how Google has accessed the contents. For example from within the app in which case we would have the user agent prefix. Or by some kind of external spider, in which case we would not have the prefix.
 

rabbitBUSH

Well-Known Member
Licensed User
Just rereading the thread raised a thought: in #1 Martin quotes Google policy
About the Sensitive Events policy
We don't allow apps that lack reasonable sensitivity towards or capitalize on a natural disaster, atrocity, conflict, death, or other tragic event.

It says 'lack reasonable sensitivity' this means their rule misses its own point and simply does a one-size-fits-all approach BANG. There can't be anything insensitive about giving the reason for closing the office. Anyway.

But let's say the Titanic sank yesterday and a client wanted an app changed to point to their life insurance provisions for survivors to know how to proceed. That line in the policy would block the app {and maybe the web sites themselves} from what in today's world is a legitimate and known comms line.

that's where it all is stupid - we call this an Erma Bombeck after the Comedienne who had a piece about one-size-fits-all panty-hose {really really funny /Erma not Google}
 
Top