Technology

CAPTCHA Has Started Asking Questions Only a Traffic Engineer Could Answer

"I AM NOT A ROBOT!"

13th August 2026

There was a time when proving you were human was simple.

A website would show you some slightly wonky letters and ask:

“Type the characters you see above.”

Easy.

K7P4M.

Done.

Humanity confirmed.

You could continue buying socks.

But apparently the robots got better at reading, because CAPTCHA has now escalated into a full psychological assessment conducted by Google’s Department of Roads, Bridges and Municipal Infrastructure.

You click:

☐ I’m not a robot

And the computer responds:

We’ll see about that.

Suddenly you’re staring at nine grainy photographs of an American intersection being asked to:

SELECT ALL SQUARES CONTAINING A TRAFFIC LIGHT

Fine.

There’s a traffic light.

Click.

Another traffic light.

Click.

But then you notice something.

In square number six, approximately three pixels of traffic light pole are visible.

Does that count?

It says traffic lights.

Not traffic-light poles.

But the pole is technically part of the traffic-light infrastructure.

What does Google want from me?

What would a robot do?

Oh God.

I’m thinking like the robot.


It Starts Innocently Enough

The first challenges are reasonable.

Select all images containing cars.

Easy.

Car.

Car.

Car.

Submit.

Then CAPTCHA decides you’re suspicious.

Please try again.

Excuse me?

I have been identifying cars since approximately 1983.

I know what a Ford Focus looks like.

But apparently one of those blurry objects in the distance was also technically a car.

Now we’re doing another test.

Select all images containing motorcycles.

Fine.

Motorcycle.

Motorcycle.

Wait.

Is that a motorcycle?

Enhance.

ENHANCE.

It’s either a Honda or a patio chair.

I genuinely can’t tell.

Meanwhile the website is silently judging you.

Interesting. A human would know.


Then Come the Traffic Lights

CAPTCHA is absolutely obsessed with traffic lights.

Nobody knows why.

Apparently the single most important skill separating humanity from artificial intelligence is our ability to identify municipal traffic-control equipment.

Select all squares with traffic lights.

You get a photograph divided into sixteen squares.

One traffic light occupies six of them.

Easy enough.

But its tiny corner extends approximately 0.4 millimetres into another square.

Do you click it?

This has become a philosophical question.

When does a traffic light stop being a traffic light?

If 98% of the square contains sky but 2% contains the corner of an amber lens, is that square with traffic light?

You hover over it.

Click.

Unclick.

Click again.

You haven’t experienced this level of uncertainty since someone asked whether you wanted to pay in pounds or euros at an airport cash machine.


CROSSWALKS

Then CAPTCHA unleashes its masterpiece.

Select all squares containing crosswalks.

Sorry.

What exactly constitutes a crosswalk?

The white stripes?

The entire crossing?

The pavement leading towards it?

Does the dropped kerb count?

What about the bit where the paint has worn away?

Also, I’m British.

It’s a zebra crossing.

Already this examination is culturally biased.

I’m now staring at a photograph of downtown Phoenix trying to understand American road markings like I’m sitting the Arizona driving theory test.

One wrong square and apparently I’m a Russian botnet.


BICYCLES ARE WORSE

Select all squares containing bicycles.

There’s a bicycle.

But the front wheel extends into the next square.

Does the wheel count as a bicycle?

Obviously yes.

But does a square containing only six spokes count as containing a bicycle?

You click it.

CAPTCHA pauses.

That pause is horrible.

You can almost hear the computer thinking.

Hmmmm.

Then:

Please try again.

YOU BASTARD.

Apparently the spokes didn’t count.

Or perhaps they did count and you missed the tiny bicycle reflected in a shop window three squares away.

Nobody knows.

There is no feedback.

CAPTCHA doesn’t explain your mistake.

It simply hands you another exam.


THE PICTURES THAT REPLACE THEMSELVES

These are particularly evil.

Select all images containing buses.

Click bus.

The square disappears.

New photograph.

Another bus.

Click.

New photograph.

Bus.

Click.

New photograph.

No bus.

Wait.

Another square has changed.

Bus.

Click.

Another bus appears.

HOW MANY BUSES ARE THERE?

You started this challenge in 2026.

It is now 2031.

Your children have grown up.

Civilisations have risen and fallen.

You’re still clicking photographs of public transport.

Eventually nothing changes.

You wait.

CAPTCHA waits.

You click Verify.

Please try again.

At this point, becoming a robot starts sounding attractive.


WHAT EXACTLY IS A BUS?

This is where CAPTCHA begins testing your classification skills.

You are shown:

A bus.

A coach.

A minibus.

A van.

Something that appears to be an airport shuttle.

A vehicle approximately twelve pixels wide.

And what might be a tram.

Select all buses.

Right.

Does a coach count?

Technically it’s designed to carry passengers.

But CAPTCHA didn’t ask for passenger-carrying vehicles.

It asked for buses.

What about a school bus?

Definitely bus.

What about a campervan?

No.

What about a Volkswagen Transporter with eight seats?

We need definitions.

Where is the marking scheme?

I would like to appeal this examination result.


FIRE HYDRANTS

This one is particularly entertaining if you don’t live in America.

Select all squares containing fire hydrants.

British person:

“A what?”

We have fire hydrants, obviously, but they’re generally underground and indicated by little yellow signs.

Americans apparently leave theirs sitting proudly on pavements like tiny metal robots.

So now you’re examining photographs of unfamiliar street furniture.

Is that a fire hydrant?

Parking meter?

Pipe?

Postbox?

Decorative bollard?

Small municipal cannon?

Click something and hope for the best.


STAIRS

At some point CAPTCHA started asking people to identify stairs.

Select all squares containing stairs.

This sounds easy.

Until you get a photograph of some brutalist building containing:

stairs,

steps,

a ramp,

railings,

a raised platform,

and shadows that vaguely resemble stairs.

Does one step count as stairs?

Grammatically, surely you need at least two.

But perhaps Google recognises “stairs” as a category rather than a quantity.

Congratulations.

You only wanted to reset your password.

You’re now debating the philosophy of architectural taxonomy.


MOTORCYCLES VS BICYCLES

This feels deliberately designed to catch people out.

You get nine blurry images.

One contains a motorcycle.

One contains a bicycle.

One contains a scooter.

One contains a moped.

One contains something apparently photographed during an earthquake.

Select motorcycles.

Does the scooter count?

Technically no.

But what does CAPTCHA think?

That’s the important distinction now.

You’re no longer identifying objects.

You’re trying to predict what the machine believes the object is.

Which raises an uncomfortable question:

Who exactly is testing whom here?


THE WORST PART: CAPTCHA SOMETIMES DOESN’T BELIEVE YOU

You complete everything perfectly.

Traffic lights.

Crosswalks.

Buses.

Bicycles.

Motorcycles.

Fire hydrants.

You’ve essentially earned a diploma in urban infrastructure.

Click Verify.

Spinning circle.

And then…

Please complete the CAPTCHA.

I DID.

I HAVE JUST COMPLETED MORE ROAD-SAFETY ASSESSMENTS THAN SOME LOCAL AUTHORITIES.

But CAPTCHA remains unconvinced.

So it gives you another one.

Select all squares containing bridges.

Bridges?!

I’M TRYING TO BUY A TOASTER.


 “I’M NOT A ROBOT” IS STARTING TO FEEL LIKE AN ACCUSATION

The checkbox itself has become insulting.

I’m not a robot.

Click.

Computer:

Prove it.

Fine.

Computer:

Identify a bicycle.

Done.

Computer:

Now a bus.

Done.

Computer:

Traffic lights.

Done.

Computer:

Motorcycles.

Done.

Computer:

Fire hydrants.

Done.

Computer:

Interesting. One more.

At some point you want to turn the interrogation around.

☐ I’M NOT A ROBOT

Okay, Google.

Your turn.

Select all squares containing:

a British pelican crossing.

Go on.

We’re waiting.


WHAT IF CAPTCHA KNOWS SOMETHING WE DON’T?

Maybe this is actually training for the future.

Perhaps society will eventually collapse and the survivors will be selected entirely according to CAPTCHA performance.

A mysterious government bunker opens.

A voice announces:

“Humanity needs people capable of rebuilding civilisation.”

Doctors step forward.

Engineers.

Scientists.

Farmers.

Electricians.

The official shakes his head.

“Sorry.”

Then he points towards me.

“You.”

Me?

“Yes. Our records show that in 2026 you successfully identified nine consecutive fire hydrants.”

My moment has arrived.


THE ROBOTS ARE PROBABLY BETTER AT THIS THAN US NOW

That’s the real irony.

CAPTCHA exists to separate humans from machines.

Yet artificial intelligence can recognise objects in photographs extremely well.

Meanwhile I’m staring at square seven saying:

“Is that a bloody bicycle?”

The robot has already identified:

  •  make
  • model
  • colour
  • approximate manufacturing year wheel diameter
  •  location
  • weather conditions
  • tyre pressure

I’m still wondering whether the handlebars technically extend into square eight.

Eventually CAPTCHA may need to reverse the system.

PROVE YOU ARE HUMAN

Select every square you’re not entirely sure about.

Human:

“Most of them.”

ACCESS GRANTED.


The CAPTCHA of the Future

Things will inevitably get harder.

By 2030 we’ll be seeing:

Select all junctions where the vehicle in lane two has right of way.

Then:

Identify every bridge constructed after 1974.

Followed by:

Select all traffic signals that comply with current Department for Transport visibility requirements.

And finally:

Calculate the optimal traffic flow through this six-lane roundabout during rush hour. Show your workings.

You have 45 seconds.

You fail.

Please try again.


I Just Wanted to Log In

And that’s what makes CAPTCHA so beautifully ridiculous.

You begin with the simplest possible objective.

Check your email.

Order some shoes.

Reset a password.

Read an article.

Five minutes later you’re conducting a forensic analysis of road infrastructure in Sacramento.

You’ve identified seventeen traffic lights, eleven bicycles, four buses, six motorcycles and something that may or may not have been a fire hydrant.

Finally:

✓ Success!

You’re human.

Congratulations.

You feel strangely proud.

Then the website loads.

Your session has expired.

Please log in again.

You enter your password.

Click Continue.

And there it is.

Waiting.

Nine photographs.

One sentence.

Select all squares containing traffic lights.

Oh, for God’s sake.

error: Content is protected !!