Skip to content
Writing
ordinary search

Everything it measures is real. It is not everything.

The gap is not accuracy. The established platforms are good at what they do. The gap is that the question a buyer asks now gets answered somewhere none of them look.

Published 2026-08-20 | 5 minute read

This is not an argument that SEO tools are wrong. They are mostly right about the things they measure, they have spent fifteen years getting there, and we do not out-depth them at it. It is an argument about the shape of the hole.

Four things none of them can see

  • Whether an assistant names you. There is no position, no impression and no session, so nothing in the instrument fires.
  • Who it names instead. Your competitive set as a model sees it is often not the set you track, and it changes.
  • Which pages the answer read. The citation list is the actionable half and it does not exist in any rank tracker.
  • Whether a change is real. Answers vary between identical asks; a tool built for rankings has no reason to carry a variance test, so it reports movement as change.

That last one is the quietest and the most expensive. In a category we measured, ordinary variation on a weighted share figure ran about five points wide across ten samples with nothing changing. Any tool reporting a four point move as an improvement is reporting noise with a direction on it.

What they are still the right tool for

Position tracking on queries that still return a list. Technical crawling. Backlink discovery at a depth nobody new is going to match. Keyword volume from a provider with a decade of data behind it. If somebody tells you those stopped mattering, check what they are selling.

The useful posture is not replacement. It is knowing which of your questions still resolve to a list, and which now resolve to a paragraph, because those need different instruments.

The joining-up problem

Running two tools that each see one surface produces two confident reports that disagree, and somebody has to pick one. The contradiction is usually where the real finding is — ranking first on a question whose answer never cites you is a specific diagnosis, and neither tool alone can state it.

questions

The short answers

What can SEO tools not measure in 2026?

Whether an assistant names you, who it names instead, which pages the answer read, and whether a difference between two readings is real. None of those produce a position, an impression or a session, so nothing in a conventional instrument fires.

Are SEO tools still worth paying for?

Yes, for what they do: position tracking on queries that still return a list, technical crawling, backlink discovery and keyword volume. Anyone telling you those stopped mattering is selling the replacement.

Why can a rank tracker not see AI visibility?

Because there is no rank. A composed answer names businesses inside a paragraph; there is no ordered list of links for the tracker to read a position from.

Do SEO tools account for how much answers vary?

Generally not, because a ranking rarely moves for no reason and a composed answer does. In a category we measured, variation on weighted share ran about five points wide across ten samples with nothing changing underneath.

Should I replace my SEO tool?

Usually not. The question is which of your queries still resolve to a list and which now resolve to a paragraph, because those need different instruments and most businesses have both.

Find out in under ten minutes

One free check, no account and no card. You get the answers the assistants gave, not a score.

Every figure quoted here comes from a real question put to a real assistant, saved with the date and the assistant that answered it. This piece was written to answer the search "seo tools limitations".

free scan

are you in the answer?

Five real questions against a live answer engine with web search on, each asked ten times. That is fifty real calls, so the report lands in under ten minutes rather than while you wait. You see every question before anything runs, and you get the answer text, who got named and the sources the engine read, not a score.

your site/what you sell/where your buyers are/the questions