The worst part about the slow animation is that when an image you've clicked is fading out, you might think you've completed everything and then click the "Verify" button at the bottom. But then that causes you to have to restart if that wasn't the last image. This is the part that convinces me that ReCAPTCHA was made to fuck with people.

I've never understood why they don't just say what you're supposed to select - it it just the lights, or the poles too? What if part only enters a box by a few pixels? Just tell me what you want, dammit!

You're not supposed to figure out the traffic light riddle, you're not a human if you do.

Yes, it's the most obvious internet rage trigger.

But I can't figure out why they make a 'delay'? Why not just show the next dam image?

To make it more expensive for bots to try this at scale, obviously. Unfortunately it's making it more expensive for humans too.

you may be right about the fade-out and delay but the time spent fading in only hampers humans not bots. As soon as it starts fading in the image is present in non-faded form and the bots can start processing it.

You realise you've just described why this would distinguish between bots and humans.

(And yes, I'm also driven to rage by slow-fade animations. A practice I can date back to Microsoft's Clippy, which, when you punched it in the fact to go away, had just one more gratuitous animation just to twist the knife that just more.)

No, it doesn't help distinguish the two, because this check can be easily circumvented by adding a small, random delay.

To reiterate: the primary goal seems to be slowing down bots.

But does it slow the bot down in a meaningful way?

If you have one IP, there's a limit on captchas solved that you're going to blow through with or without the delay.

If you have a bunch of IPs, you can multithread the solving.

> the time spent fading in only hampers humans not bots

Not necessarily, contrast adds detail and mistakes are expensive, so bots too are incentivized to wait for the final picture (this assuming that network communications aren't monitored to get the incoming image out of the request).

Also clicking on that image too early is a good signal that it's a bot.

The bot presumably is running in something like chrome headless or selenium (if you're processing JS), so it would have access to the image the moment the response is received.

Unless Google is literally streaming in the image frame-by-frame, I'll admit I haven't looked into the details but this doesn't seem likely as it's pretty complicated compared to just using an image.

The bot would just read the unfaded image from the DOM.

> Clicking the image too early is a good signal that it’s a bot.

The fade in is actually a nice gesture to the human to show them that an image will be there soon, while still slowing them down to rate limit the bots.

I don't agree that's it.

... it really doesn't make it that much more expensive for bots, it's just a short delay. In fact, I doubt it makes a difference at all.

But it makes things really annoying for humans.

So I don't see any advantage in that trade-off.

Include the traffic light poles.

I've never included the traffic light poles.

If I don't include them, I get asked more. (Or occasionally 'select ALL the...'.) If I include them, it usually goes away.

I don't include the poles, how about wires? What about pedestrian lights, are they "traffic lights"?