My $ amount has remained the same since 1 PM MST, impressions have climbed, and RPM has gone down 80 cents in that time period. Doesn't make any sense.

Now they're in the process of taking money away from me! This is a joke.

normal # of clicks, very low Ad views reported + low revenue...

There's probably not even much sense in discussing it anymore. It is what it is, and what it is is not good.

7 PM now and still no movement.

Mine made a leap in views and RPM, and I got a few more dollars, and now they've taken it all away again.

So far today a total of eleven clicks taken back that I have been able to track. There are 3 more suspected take-backs but I'm only counting the different "totals". The peak was 11 clicks and since then I've had 6, 7, 8, 10 and 7.

However the 4 clicks on my mobile "style" have stuck around since 8am this morning reinforcing my belief that Google does not think mobile devices play a big part in click-fraud schemes.

Two more gone... now a total of 13 clicks taken back. But the 4 mobile clicks are still there...

@desert wa rat

This kind of makes sense (your mobile observation) because if a site is only serving mobile ads to a mobile device then either the bot has to present itself as a mobile device or actually be running on a mobile device (anyway, I think that would be the case). Maybe tackling the mobile side of things presents issues for these fraudsters that make it more risky or just more complicated when they're really out for the low hanging fruit on the tree.

@Webcentric Yes... I think that's the key. Plus the smartphones and tablets run Linux or a variant of Unix (the Mach kernel in the case of IOS). Non jailbroken IOS devices are more difficult to exploit. The jury is still out for Android. But the fact that these devices change IP address as they move from cell to cell could make it difficult for the C&C Servers to maintain contact.

Despite the contrary propaganda, it is much more difficult to exploit Linux/Unix than it is Windows. So a Linux desktop on your web site is also less likely to be an infected click-bot.

Not to say that a clever design couldn't spoof its host. One reason I run Linux as a desktop is that I've had dozens of Linux/Unix computers with static IP addresses on the Internet with nothing but a Cisco router between them and the masses and have had exactly two of them exploited since 1994. :)

Before blaming bots, make 'absolutly' sure that any tooltip (or social plugin) window doesn't float on your ads. Make sure ads are not blended completely and difficult to distinguish from the content.

If your ads are too close to clickable elements, too much blended inside content, covered by floating windows, Google flags clicks, page or even the account.

There were a couple of threads in Adsense support forums, filled with people saying that 100% of their text-ads clicks get shaved, and Adsense support responding that the issue is with their placements.

Before blaming bots, make 'absolutly' sure that any tooltip (or social plugin) window doesn't float on your ads. Make sure ads are not blended completely and difficult to distinguish from the content.

We've covered that (in the February thread). I can't see how my ads are overly susceptable to a sudden onslaught of accidental clicks. Just a phpbb3 forum with one banner at the top and one at the bottom. Nothing fancy and they haven't changed in 12 months.

And we're now up to 16 clicks taken back today even though traffic and page views are up.

The idea is even if you didn't change something, Google probably did change its definition of invalid clicks. And they indeed have changed their ad placement policies and, reportedly, their algorithms.

The idea is even if you didn't change something, Google probably did change its definition of invalid clicks. And they indeed have changed their ad placement policies and, reportedly, their algorithms.

That seems to me to be quite clear. Something changed around the first of the year. I wonder if Google's aquisition of that anti-click-fraud outfit has anything to do with it.

I I know is that yesterday was stuck and never recovered. The worst fay this month. Today impressions climb just like yesterday and CTR is 0. Looking like a mirror day of yesterday. Everyone is talking about clicks being shaved off. What I'm seeing now is not enough clicks to take away.

I only wish mine were stuck. Up and down literally every five minutes. Best not to check the same day anymore, but wait a day or two, at least for my accounts.

I have 1000 visitors so far today and not a single click. Is there something dodgy going on here?

I have 1000 visitors so far today and not a single click. Is there something dodgy going on here?

Yes! I'm having the same issue. Impressions continue to climb and not a single click, only PPI pennies.

This is seriously messed up! It's been over 24 hours with not one click! With over 7000 impressions? What ever you say G.

Earnings yesterday were all over the map, up and down like a yo-yo. I have chat option so I talked to them this morning. She asked if I'd changed my ad placement. I said no. Nothing has changed. She said it sounds like "global sabotage" and to check my log files and let them know if I find anything suspicious.

The only clicks G is not taking back are on my mobile ad. Steady as a rock; both CPC and earnings. Just not enough of them, darn it. But at least G isn't snatching them away.

She said it sounds like "global sabotage" and to check my log files and let them know if I find anything suspicious.

I'm not sure what she meant by "global sabotage". I'm pretty sure this is happening to everyone, big or small, just that many don't check the performance reports very often and don't notice. Start checking them every 10 or 15 minutes and you'll see clicks appearing and disappearing.

Is a "suspicious" user one who qualifies as a bounce? Did she offer up any criteria?

She said to look for IP addresses that come once an hour or come and stay for a second or two or generate an inordinate number of pageviews. I can see pageviews by day, but I don't know how to see if an address comes once a day or just stays for a very short time.

You have to either pull your log files into an analyzer (or Excel) or maybe use StatCounter or Piwik to get the IP numbers.

Thanks, netmeg. I forgot about statcounter.com

She said to look for IP addresses that come once an hour or come and stay for a second or two or generate an inordinate number of pageviews. I can see pageviews by day, but I don't know how to see if an address comes once a day or just stays for a very short time.

That's interesting... I found a user who conforms to that description. Piwik's log files of visitors will actually give a "profile" of a user and one person has visited my site 41 times in 50 days but only spends a few seconds while reading 3 or 4 pages: "Spent a total of 33 min 58s on the website, and viewed 143 pages in 41 visits.
Converted 0 Goals."

The idea is even if you didn't change something, Google probably did change its definition of invalid clicks

That is what this is starting to feel like. I'd really first like to know if this is the new normal.

"?..the new normal"? Probably not. This is a work in progress. We're seeing ad hoc solutions being layered over an existing system. I'm fairly certain that eventually these new compenents will find their way into the main processing algorithm and become invisible to us. They'll continue to do their work but we won't see the yoyo any more. We seem to be in a phase where it's easier for G to run processes independently than all together.

I think Webcentric is right. Google (or the company Google bought that specializes in click-fraud) is trying out ideas. Although, frankly, I would have thought they'd have set up a honey-farm and got it all right before implementing it. But the facts seem to point to experiments that cause Analytics to disagree with Reporting (for instance).

But we will almost certainly continue to see some yo-yo effects simply because the new breed of click-fraud bots are impossible to block and extremely difficult to detect.

I'm starting to see some patterns, however. Clicks seem to be more likely to be taken back if:

1) They are high-bid;
2) They show up in a burst;
3) They are not from a mobile device.