Your Rank Tracker Is Lying To You, And Your PageSpeed Score Is In On It

Everybody in marketing says the same thing with a straight face. Watch your metrics, trust the data, let the numbers guide the decisions. It sounds responsible. It sounds grown up. And most of the time it is a quiet way to avoid thinking.

I want to push on that a little today, because a few recent stories in the search world all rhyme in a way that should bother you. Rank tracking tools have been spitting out weird results, sometimes a single word where a keyword should be, sometimes a redirect URL instead of the real page. PageSpeed scores keep getting treated as reality when they are closer to a weather forecast. And a shared way to tell AI companies not to train on your content just showed up, which forces a decision most people would rather not make.

Different headlines. Same lesson. We have fallen in love with our instruments and forgotten what they actually point at.

The number that feels like truth

Here is the thing about a clean dashboard. It feels like knowledge. A green score, a position number, a tidy line going up. Your brain reads it as fact because it arrived with decimal points and a chart behind it.

But a metric is not the world. A metric is a photograph of the world, taken through a specific lens, at a specific moment, by a tool that had to make a hundred small assumptions to get the shot. Change the lens and the picture changes. That does not mean the picture is useless. It means the picture is a picture.

A metric is not the world. It is a photograph of the world, taken through one lens, at one moment, by a tool making a hundred quiet assumptions.

We forget this constantly. We treat the ranking number like it is the position. We treat the speed score like it is the experience. And then we optimize toward the number instead of the thing the number was supposed to represent. That is where the trouble starts.

When your rank tracker reports nonsense

Let me be specific. Rank tracking tools pull their data by imitating a search. They send a request, read what comes back, and parse out where your site sat. That is a fragile little dance, and when the search results pages change shape, the dance stumbles.

So you get strange output. A keyword that resolves to a single word. A tracked position that points at a redirect URL rather than the page people actually land on. The tool is not broken in some dramatic way. It is doing exactly what it was built to do, reading a page that no longer looks the way it did when the parser was written.

Now picture the marketer who does not know that. They open the report, see rankings slip or scramble, and panic. They file a ticket. They rewrite a page that was fine. They tell the client something is wrong when nothing is wrong except the instrument. I have watched it happen. A whole afternoon burned chasing a ghost in a spreadsheet.

The fix is not to stop using rank trackers. They are useful. The fix is to hold them at the right distance.

A rank tracker tells you a story about your visibility. It does not tell you the truth about your visibility. Those are different claims, and the gap between them is exactly where good judgment lives. When a number looks insane, your first move should be suspicion of the number, not panic about the business. Does the tracked URL even match the page you care about? Is the result a redirect? Did the layout of the results page shift last week? Ask that before you touch anything.

Lab coats and lab scores

PageSpeed numbers have the same disease, just dressed nicer.

Here is what a lot of people miss. A lab score is run in a controlled environment. A simulated device, a throttled connection, a clean room with no variables. It is a laboratory test. It is repeatable and tidy and it will give you the same answer twice, which feels wonderful and proves almost nothing about your Tuesday afternoon visitor on a three year old phone sitting on a train with two bars of signal.

That second thing is field data. Real people, real devices, real networks. And field data is messy, because real life is messy.

So you have teams celebrating a lab score in the nineties while actual humans wait and bounce. Or the opposite, teams terrified of a mediocre lab score while their real users have a perfectly fine experience. Both groups are reading the wrong gauge and acting on it with total confidence.

A lab score tells you what is possible. Field data tells you what is happening. Confusing the two is how you win an argument and lose a customer.

I am not anti speed. Speed matters enormously, and I will fight anyone who treats a slow site as acceptable. But a hammer is not a house. A fast lab score is a tool you use to build a fast experience. It is not the experience. If you chase the score for its own sake, you can absolutely game a lab number while the thing that number was invented to protect, a real person getting what they came for quickly, quietly gets worse.

Okay, so what do you do with that? You treat the lab score as a diagnostic, a place to find problems, and you treat field data as the verdict. The lab says here is where it hurts. The field says here is whether anyone actually felt it. You need both, and you need to know which one is which.

The map is not the territory

There is an old idea worth dragging into this conversation. The map is not the territory. A map is a useful, deliberate simplification. It leaves things out on purpose so you can see the roads. But if you mistake the map for the land, you will eventually drive into a lake.

Every marketing metric is a map. Rankings are a map of visibility. PageSpeed is a map of performance. GA4 sessions are a map of behavior. Each one is drawn at a certain scale, with certain roads included and certain fields left blank.

GA4 deserves a specific mention here, because it tripped up half the industry when it landed. People assumed the numbers meant the same thing as the old numbers. They did not. The model changed underneath the labels. A session in GA4 is not counted the way a session used to be. So reports that looked like declines were often just a different map of the same territory, and plenty of smart people made real decisions based on a comparison that was never valid in the first place.

The skill nobody puts on their resume is knowing what each number actually measures and, just as important, what it quietly leaves out. That is not a glamorous skill. It will never trend. But it is the difference between a marketer who reads data and a marketer who is read by it.

The opt out nobody wants to decide on

Now the newest wrinkle, and the one I find genuinely interesting. There is a growing effort to give publishers a shared, standard way to tell AI companies not to train their models on your content.

On the surface that is a tidy technical story. A signal, a setting, a line in a file. But underneath it is a strategy question that most businesses have not even noticed they need to answer, and I think that is a mistake.

Think about what is actually being asked. Do you want your content used to train the systems that more and more people now ask instead of searching? Because there is a real tension here, and it is not obvious which side you should be on.

Say yes, let them train on everything, and you lose some control over how your work is absorbed and reused, possibly with no credit and no click back to you. Say no, lock it all down, and you may remove yourself from the very systems that are starting to recommend brands to buyers who will never type your name into a search bar. You could protect your content right out of the conversation.

That matters.

I am not going to pretend there is one correct answer, because there is not, and anyone who tells you otherwise is selling something. A reference publisher whose whole value is being the source might reasonably want to be everywhere an AI can reach. A business whose edge is proprietary insight, original research, a point of view competitors cannot copy, might reasonably want to guard the good stuff and feed the machine only the surface.

But here is my actual point, and it ties back to everything above. This is a decision, not a default. The worst outcome is not choosing wrong. The worst outcome is letting a setting decide your AI strategy because nobody in the building understood that a strategy question was hiding inside a technical toggle. Same disease again. We let the instrument make the call because we never asked what the call was really about.

Being recommended is the new being ranked

Let me connect the opt out to where visibility is actually heading, because this is not abstract.

For twenty years the game was ranking. Get the blue link higher. Now a growing slice of people ask an AI tool a question and take the answer, or ask for a recommendation and take the shortlist. They never see ten links. They see a sentence, maybe a name or two.

So the question shifts from where do I rank to am I in the answer. And whatever metric eventually claims to measure that, treat it with exactly the suspicion we have been building this whole time. The early tools that try to track AI visibility are going to be shaky, inconsistent, and easy to misread, in the same way rank trackers stumble when the page changes shape. The AI answer is not a fixed thing you can screenshot and trust forever. It moves. It varies by phrasing. It varies by user. Any dashboard that promises a clean number for it is handing you a map of fog.

Which means the old discipline matters more, not less. Do not fall in love with the proxy. Keep asking what the number is standing in for, and whether the number and the thing have quietly drifted apart.

Trust the thing, not the gauge

So let me pull this together, because the three stories really are one story.

Rank data can report a single word or a redirect, and people rebuild healthy pages because of it. Lab speed scores shine while real users wait, and teams celebrate the wrong win. An AI training opt out shows up, and businesses risk sleepwalking through a strategy decision because it arrived disguised as a setting.

In all three, the instrument got mistaken for the reality it was built to describe. That is the mistake. Not the tools. The tools are fine. The mistake is the quiet, comfortable laziness of trusting the gauge and never checking the engine.

Fall in love with the question your metric is trying to answer. Never fall in love with the metric.

Measure everything you can. Rankings, speed, behavior, AI visibility, all of it. Then do the harder thing, the thing the dashboards cannot do for you. Ask what each number actually sees, what it is blind to, and whether it still points where it claims to point. The marketers who keep winning as search splinters into a dozen answer boxes will not be the ones with the prettiest dashboards.

They will be the ones who never confused the photograph for the face.