[sly010]

The examples in the article are all kind of acceptable, but I have asked google some very factual question (e.g. can covariance be negative) where it picked out the wrong answer from stack overflow that was clearly the opposite of the truth. It confused the hell out of me.

[tyingq]

Google sometimes gives out pretty dangerous advice as well: http://imgur.com/a/bsP98

[rchaud]

It's frankly quite shocking that Google's ad results are displayed with pretty much the same look and feel of their search results. Yes, the more tech-savvy among us will see the "Ad" button in gold, and those with ad-blockers won't see the ads at all.

But just a few years ago, Google's ads would appear in a sidebar, clearly separated from the search results content.

Not everyone is able to tell the difference today, and I think Google exploits this. I recently activated a new B2B website for an client and the senior manager was an older gentleman in his 50s. He never uses the browser address bar to navigate to websites, preferring instead to use the Google search box on his homepage. He sent me a panicked email saying when he typed in the new website's name, the Google search results showed competitors' names ahead of his. This was well after Google had crawled the site and it was showing up on the results page.

As it turned out, he did not understand that the first few results he was seeing were ads. While he isn't a savvy user, I would wager that many, many people are similar to him in that they don't exercise good judgement when they're on the web and would easily be fooled into thinking that something like "proton therapy" is a cure for cancer because it's at the top of the search results.

[djsumdog]

I run uBlock Origin, so when I see an ad it tends to really stand out (like on the site this article is on, which has an ad that doesn't get blocked currently).

That screenshot is pretty shocking. That's an insane amount of pixel space dedicated to the ads, then the Google featured snippet and finally the actual search.

[thomasahle]

They always (at least since 2007) had ads in the main search result column: http://searchengineland.com/search-ad-labeling-history-googl... They've certainly removed some of the distinguishing features though.

[tyingq]

Here's a pre-2007 result. It's still at the top, but across both columns.

http://static1.businessinsider.com/image/4f2992c7eab8ea28690...

The point that they've slowly made changes over the years to make the ads look less like ads is very clear though.

They've also generally done a lot of work to push the organic results down. It's very easy to find queries where there's no organic result showing above the fold, even on a fairly decent desktop monitor.

[mohn]

There aren't a lot of clinics that can do it, but proton therapy has emerged as another method for treating tumors by irradiating them. Unlike a beam of high energy photons (used in conventional radiotherapy), a proton beam can be tuned so that most of its radiation dose will be deposited at a specific depth. By contrast, a photon beam will deposit more energy at depths shallower than the target, and it will also deposit some energy at depths beyond the target.

https://physics.stackexchange.com/questions/169665/dose-dept...

[cflewis]

I'm a Googler. I've reported this internally.

[bdrool]

That's all well and good, but it doesn't really matter if individual cases get squashed. The larger issue is that Google has moved away from the stance of "we just give you what you search for, it's up to you to verify it" and is instead presenting some results as "this is the definitive answer". That's bad even if all the horribly wrong results are removed. Does anyone really think it'd be a good idea for a single company to own what is and is not considered "true"?

[mohn]

Good screenshot, that is a very unfortunate snippet. It's interesting to me that the quackery text in the snippet closely matches the input query about carrot juice, but the top ad result (with very little text similarity) is for a cancer center with a real radiotherapy treatment [1], FDA approved since 1988.

Proton beam therapy is not widely known, which might explain the scare quotes in rchaud's comment suggesting that non-savvy Google users "would easily be fooled into thinking that something like 'proton therapy' is a cure for cancer because it's at the top of the search results".

I detest ads, so it feels weird to write something like praise for this one. I guess I dislike quackery even more. Now I want to try similar search queries to find other cases where a Google ad suggests a real treatment amid all the bullshit in the non-ad search results.

[1] https://en.wikipedia.org/wiki/Proton_therapy

[folli]

That's actually a really scary example.

[TheOtherHobbes]

It also highlights why ML graphing techniques are a poor (lazy?) way to represent knowledge.

Having a data structure with an entry in it is not the same as knowing something with an acceptable level of confidence.

This is basically automated Chinese whispers. There's no sanity checking, no peer review, and no reality testing.

[tyingq]

It's not the worst one.

Here's one where absolutely NOTHING above the fold is a good idea. http://imgur.com/a/RXABg

[lordCarbonFiber]

That's a really interesting problem space, because what do you do when the foundation of the question is terrible, and yet, there exists plenty of information saying otherwise. What is even the ideal result from that query?

Play stupid games, win stupid prizes I suppose.

[tyingq]

I agree in terms of the organic results that come back.

However, I don't see the logic in picking one of them and highlighting it. It has the appearance of an endorsement.

I would also argue that the ads are deceptive for this query. Surely there's some AdWords policy about that?

[thomasahle]

Google always did this: http://imgur.com/1ibgofn Only the source used to be above the snippet rather than below it.

I guess the main problem is with voice search, since you only get the one result.