Rank trackers disagree and which number to trust

The screenshot that starts every argument

I’ve had the same conversation with three different site managers this year. Someone opens a rank tracker, sees position 4 for a keyword, then opens a second tool and sees position 9 for the same keyword on the same day. They ask me which one is lying. Neither is lying. They’re both telling the truth about a search that doesn’t exist anymore, because the search that actually matters is the one your specific visitor ran, on their device, in their location, logged into their own Google account, at their own moment in time. No tracker simulates that. They all simulate something close to it, and “close” is where the disagreement lives.

I run a portfolio of sites and I’ve stopped treating any single tracker’s number as gospel, mostly because I got burned trusting one too literally. I want to walk through the actual mechanics of why these tools diverge, because once you understand the mechanics the disagreement stops being confusing and starts being useful.

Different data centers run different searches

Every rank tracking tool works the same basic way: it sends an automated search query from a server, usually through a proxy or a cloud data center, and records where your URL lands in the results. That server has an IP address tied to a location. Google uses that location signal, along with the language and region settings you configured for the check, to decide what results to show.

If tool A checks from a data center in Virginia and tool B checks from one in Frankfurt, and you asked both to check “rank tracking tools” with a US, English targeting, they should land close together. But “close” isn’t “identical,” because Google doesn’t serve the exact same index snapshot to every data center at every moment. Index updates and ranking adjustments roll out gradually, not as a single global flip of a switch. A tracker that happened to query during a rollout window can catch a different state of the results than one that queried an hour earlier or later.

This is also why a tracker set to “United States” isn’t automatically checking from where your actual customers are searching. If most of your traffic comes from Texas and your tracker’s default is a generic US location, you’re comparing your real-world performance against a proxy for an average that may not represent any single one of your visitors.

Desktop and mobile are not a rounding error

Google has run mobile-first indexing for years, and mobile and desktop result pages are frequently not the same list in the same order. Local packs, featured snippets, People Also Ask blocks, and shopping carousels all behave differently by device, and each one of those elements can bump your organic listing up or down a slot without your page changing at all.

If tool A defaults to desktop and tool B defaults to mobile, you will see two different numbers for the same keyword, and both will be accurate readings of two different SERPs. Before you compare tools, check what device each one is actually querying as. This single setting explains a huge share of the “why don’t these agree” tickets I get.

Personalization you can’t fully strip out

Rank trackers try to check results in a clean, logged-out, no-history state, because that’s the only way to get a repeatable number. But Google still uses signals a logged-out automated query can’t fully erase: the exact data center handling the request, recent query patterns from that IP range, and small experiment buckets Google runs on live traffic. Two tools querying from two different proxy networks can land in two different experiment buckets without either one doing anything wrong.

What this means practically: the number in your tracker is not “where you rank.” It’s “where a clean, geographically approximated, logged-out version of you ranks, at the moment the tool happened to check.” That’s a useful proxy. It is not the same thing as what your actual next visitor sees.

Sampling frequency hides or reveals volatility

Some tools check once a day. Some check multiple times a day. Some let you set the frequency yourself. If your keyword sits in a spot that swings between position 5 and position 11 over the course of a day, and one tool only samples at 3am while another samples at noon, you will get two “official” numbers that both happened, just at different hours.

I’ve watched this play out directly on money keywords for sites in my portfolio. A once-daily check can make a page look rock solid at position 6 for a month, while an hourly check on the same page shows it bouncing between 4 and 14 constantly. The daily number wasn’t wrong. It just threw away information the hourly check kept. If a keyword matters enough to make decisions on, check its sampling frequency before you trust its trend line.

SERP features eat the position you’re tracking

Rank trackers count organic blue links, but the page a real person sees is stacked with things that aren’t organic blue links: ads, a local pack, a featured snippet, video carousels, image blocks, People Also Ask. Some tools count position 1 as the first organic result regardless of what’s above it. Others count pixel position, meaning if there are three ads and a local pack above the fold, your “position 1” organic listing might be the eighth thing on the actual page.

This is the single biggest source of tools disagreeing on a keyword that hasn’t otherwise changed. Nothing happened to your rankings. A SERP feature appeared or disappeared, and one tool’s counting method reacted to it while another tool’s method didn’t.

Search Console isn’t the neutral referee either

The instinct once you’ve been burned by tracker disagreement is to fall back on Google Search Console’s “average position,” treating it as the real number since it comes straight from Google. It’s more honest in one sense, because it’s built from real impressions instead of simulated searches. But it has its own distortions. It’s an average across every query variant, every device, every country, and every SERP feature state that triggered an impression for that page, blended into one number. A page that ranks 2 for a branded variant and 15 for the head term can average out to something that doesn’t describe either search accurately. GSC also samples and rounds, and its numbers for low-volume queries can be noisy in ways that look like movement but aren’t.

So GSC isn’t the tiebreaker that makes tracker disagreement go away. It’s a third data source with its own blend of accuracy and blind spots, useful specifically because its distortions are different from a rank tracker’s distortions.

What I actually do with the numbers

I pick one tracker as my primary reference for a given site, keep its settings (location, device, frequency) fixed, and judge trend, not absolute position. If a page moves from 14 to 6 over three weeks on that one consistent tool, I trust that direction regardless of what a second tool says the exact number is. I cross-check against Search Console impressions and clicks for the same query, because a position improving on a tracker that isn’t showing up in more impressions or clicks in GSC tells me the tracker’s location or device setting has drifted from where my actual audience searches. And I stopped comparing raw numbers across tools entirely. If someone asks me “why does tool A say 4 and tool B say 9,” my answer is usually to check whether they’re set to the same device and location before assuming either tool is broken.

None of this tells you how long it takes to move a number, and I’m not going to pretend I know Google’s internal logic well enough to promise a timeline or a guaranteed position. What I can tell you, from running these checks daily across a portfolio of sites, is that the disagreement between trackers is rarely a bug. It’s usually the tool telling you exactly what it’s configured to tell you, and the fix is understanding the configuration, not switching tools again.

If you want a hand making sense of your own rank data, or building out the keyword research and technical SEO underneath it, The SEO Desk can help.