Why Two Rank Trackers Report Different Positions

Because they asked slightly different questions and then counted the answer differently. Two trackers pointed at the same keyword on the same morning are issuing two separate requests, from two locations, with two sets of parameters, and applying two sets of rules about what counts as a result.

None of that makes either one wrong. It does mean “which tracker is accurate?” is the wrong question, and the useful one is “which one is consistent, and do I know what it’s consistent about?”

The request is never identical

Every tracker builds a search request out of a handful of choices, and each choice moves the result page.

Location. The biggest single source of divergence. One tracker might default to a country, another to a city, and a third to a specific coordinate. For any query with local intent — and Google treats far more queries as locally intended than people expect — those three produce genuinely different pages, not noisy versions of one page.

Language and country parameters. Interface language and the country edition of the search page are separate settings, and tools set them differently. A Spanish-language request from a US location is a different result set from an English request from Spain.

Device. Mobile and desktop are separate result sets with separate layouts and often separate orderings. If one tracker defaults to mobile and the other to desktop, the disagreement is structural rather than accidental. That one is worth its own treatment — see mobile and desktop are two rank sets.

Time of day. Rankings are not recomputed on a schedule you can see, but the page you get at 04:00 and at 14:00 can differ, and trackers run their queues at whatever hour suits their infrastructure — which is a property of how the data gets collected at all, covered in where a rank tracker’s data comes from.

The counting rules differ too

Once a page comes back, someone has to decide what “position 4” means. There is no standard, and the choices are all defensible.

Which blocks count as organic. Does a featured snippet occupy position one, or is it a feature sitting above position one? Does a local pack consume a slot? Do video carousels, “people also ask” blocks, and product grids get counted? Two trackers with different answers can report 3 and 6 for the same URL on the same page, both honestly.

Deduplication within a domain. If a domain appears twice, some tools record only the best position, some record both, and some collapse indented sub-results into their parent. When your site holds two slots, this alone explains a gap.

Subdomain and path matching. Whether blog.example.com counts as a ranking for example.com is a configuration decision, and defaults vary.

How deep it looks. A tracker that fetches two pages of results reports “not found” where one that fetches five reports position 47.

What this means in practice

Absolute position numbers are not portable between tools. You cannot take last quarter’s numbers from one tracker and this quarter’s from another and draw a line between them; the step change in the chart is the tool switch, not your SEO.

What is portable is direction and magnitude within a consistent series. If both trackers show the same keyword sliding over six weeks, that’s a real signal, independent of which one says 8 and which says 11.

This is also why a tracker’s value is mostly in its consistency. A tool that reports a systematically pessimistic number every single day is more useful than one that is occasionally exactly right, because the first one has an interpretable baseline and the second doesn’t. There’s more on that in baselines before you read a position.

When a gap is actually a problem

Most disagreement is configuration. A few kinds are worth chasing:

• The gap changes size week to week with your settings unchanged. That points at instability in one tool’s sampling rather than a settings mismatch.
• One tool reports “not ranking” while the other reports a solid page-one position. Usually a domain-matching rule or a redirect the tools resolve differently.
• One tool’s series has visible flat spots. Flat means “we reused yesterday’s value”, which some tools do silently when a fetch fails.

What to actually do

1. Pick one tracker as the system of record and treat every other number as a cross-check. Two systems of record is zero systems of record.
2. Write down its settings — location, device, language, search depth — next to the report. When someone questions a number in four months, this is the answer.
3. Compare tools once, deliberately, on a sample of maybe twenty keywords, and characterise the offset rather than trying to eliminate it.
4. Never mix series. If you change tools, start a new chart and annotate the switch.
5. Judge a tool on stability, not on matching your intuition. Your intuition comes from your own browser, which is the least neutral rank checker you own — here’s why.