Rank tracking is getting plenty of attention on LinkedIn at the moment. Thanks to Marco Janck and Dmytro Sokhach for their posts and observations, which helped refine my thoughts and come up with this article. They raise valid questions about collecting reliable search data and what happens to SEO when that becomes harder (not impossible as some might think at the moment). Their posts also got me thinking about a slightly different question: did we expect too much from this data in the first place?
There are genuine collection problems. Providers are working on fixes, while parts of the discussion have already moved on to what we will all do when the tools stop working. That is quite a journey from “we have a data problem”. Before we reorganise the profession around it, perhaps we should examine what the numbers could tell us before any of this started.
A recorded ranking was never what every customer saw. Estimated traffic was never access to a competitor’s analytics. A visibility index was never a complete measure of commercial success. Google’s changes make those distinctions more urgent, but they did not invent them.
The collection problems are real
Google’s /goto links complicate the identification of destination URLs. DataForSEO’s documentation illustrates how this can go wrong: it warns that wrapped results can contain Google’s redirect domain where someone might expect the destination website. Further down the same page, its batch-processing example identifies results using that domain field and saves a missing match as None. A result could therefore be present but missed by the code interpreting it. See DataForSEO’s documentation
That does not show how every commercial tracker behaves. It does show a plausible route from a successful request to a false “not found”. Once that reaches a report, it can look like a ranking loss rather than a problem recording the ranking.
Then there are the apparently valid but abnormal results now called “poisoned SERPs”. On 22 October 2026, Ahrefs acknowledged affected results and said it was filtering abnormal responses and cleaning up historical data. A failed request leaves an obvious gap; a misleading response can produce a convincing graph. Providers need to explain what is affected and what has been corrected. See Ahrefs’ statement
These are serious issues. They still leave some distance between unreliable observations and the permanent death of rank tracking.
We have been warned about the numbers before
In May 2022, Ahrefs compared its monthly US organic traffic estimates with Search Console data for 1,635 websites and reported a median deviation of 49.52%. That was a historical, vendor-run study, not a current accuracy score. Nevertheless, substantial differences between estimates and recorded clicks existed years before today’s collection problems. Ahrefs’ traffic-estimation study
In 2023, Ahrefs explained that changes to its keyword database and search-volume calculations had moved estimated traffic figures. Its advice included displaying database-update annotations when interpreting those movements. Before explaining why a website had apparently declined, you needed to check whether the movement came from a change in the tool. Ahrefs’ database update explanation
Rankings have a different limitation. Google explains that results can vary with time, location, language, device and personalisation. A tracker can accurately record a position under particular conditions without establishing what every potential customer saw. Likewise, SISTRIX defines its visibility index using a selected keyword set and weighting method. Both can be useful, provided we remember what they measure. Google’s explanation and SISTRIX’s methodology
Much of this has been documented for years, so blaming the vendors for every misunderstanding would be convenient but unfair. The awkward question is how an acknowledged estimate becomes a confident statement in a presentation once its qualifications have disappeared.
“Directional” still has to mean something
Third-party search data should increasingly be treated as directional evidence. But that cannot become a polite excuse for using figures we no longer understand.
A ranking observation, a modelled traffic estimate and a corrupted response require different treatment. The observation has a scope, the estimate has uncertainty, and the corrupted response may need to be discarded. Combining unreliable observations into a bigger chart does not automatically make them dependable. Two tools agreeing is not necessarily independent confirmation either, particularly if they share a provider.
The practical question is what the evidence can reasonably help us do. Third-party tools and Search Console contribute different things, so let’s separate them.
Third-party tools: what is still useful?
Third-party research gives us a view beyond our own website. It helps identify competing pages, unfamiliar terminology and customer questions we have not addressed. Uncertainty about an exact traffic figure does not remove that value, although we still need credible underlying observations.
| Task | What was never a safe assumption | What current problems put at risk | What remains useful |
|---|---|---|---|
| Track rankings | One position describes what every customer sees. | Apparent drops or missing URLs may be collection failures. | Flag persistent changes in important queries, then look for supporting evidence. |
| Researching competitors | Estimated traffic reveals actual visits or commercial success. | Faulty rankings can distort traffic estimates and apparent winners. | Find competing pages and offers worth examining. |
| Comparing visibility | An index measures the whole market or proves business improvement. | Collection gaps and dataset changes can undermine comparisons. | Follow relevant topics using consistent settings and record changes to the data. |
| Planning content | Search volume guarantees demand, visits or sales. | Suspect results can mislead gap and intent analysis. | Discover questions and page types, then check them against customer needs and credible results. |
Suppose several competitors have useful pages explaining a purchasing decision your website barely addresses. Read them and understand the problem. You do not need a perfectly accurate traffic estimate before deciding that the subject really deserves attention.
You do need to resist treating an export as a publishing instruction. In From Confirmation Bias to Creation Bias, I wrote about how available search data can start determining what we consider worth creating. Apparent gaps deserve investigation, but so do customer questions the tools barely register. Two hundred spreadsheet rows are still not a convincing reason to commission 200 articles.
GSC: useful evidence, with its own limits
Search Console should not become the next unquestioned authority simply because third-party collection is becoming troublesome. Google documents omissions and limits in its performance data, including queries excluded for privacy. These limitations exist independently of the current rank-tracking problems. Google’s explanation of Search Console data limits
The middle column below concerns what needs checking before acting. It does not mean these metrics have suddenly stopped working.
| GSC metric or task | What was never a safe assumption | What needs checking | What remains useful |
|---|---|---|---|
| Average position | A worsening site-wide average means established rankings have fallen. | Changes in the queries, pages, countries and devices receiving impressions. | Compare important pages and query groups under consistent conditions. |
| Impressions | More impressions automatically mean more valuable demand. | Which searches and pages contributed the change, and their commercial relevance. | Investigate relevant gains and losses in search presence. |
| Click-through rate | Low CTR proves the title needs rewriting. | Position, query mix, device and surrounding search results. | Identify specific listings or pages that deserve investigation. |
| Query research | The table contains every query or describes the whole market. | Privacy omissions, export limits and topics where you have no presence. | Understand language associated with existing visibility and explore relevant opportunities. |
| Clicks | More clicks prove better business results. | Which pages gained clicks and what relevant visitors subsequently did. | Track search arrivals alongside meaningful actions, where measurement permits. |
Google’s definitions explain why impressions, clicks and position need careful interpretation. Its practical guidance also shows how page and query filters help investigate changes. See GSC metric definitions and working with performance reports
Average position gives us an easy example. Imagine an established group of searches produces 100 impressions at an average position of three. You gain another 100 impressions from different searches at an average position of 23. The combined average position is now 13, although the original group has not moved. That illustrative calculation shows why a headline average cannot, by itself, establish a ranking collapse.
The same applies to CTR. Before rewriting a title, establish which queries are involved and inspect relevant results. A page gaining impressions for less relevant searches can produce a different CTR without anything changing in its listing. The site-wide average will not explain that for you.
Investigate before you start fixing things
Let’s assume a tracker reports that an important service page has lost visibility. Check whether the expected URL is still being recognised, whether the change persists and whether the provider has reported collection problems. Then examine relevant GSC page and query data, keeping country, device and comparison periods consistent, and inspect a credible sample of results.
You might find a genuine loss, a collection failure, a change in query mix or several things happening together. Stable GSC figures would not prove that every tracked ranking was unchanged, but they would give you a reason to question a claim of wholesale collapse.
The cost of the proposed action matters too. An uncertain signal can justify ten minutes of checking. Rewriting a successful commercial page, abandoning an important topic or changing a substantial budget needs far stronger reasoning. We should be able to explain that without making the client feel that all the data is useless.
Better explanations would be a start
A useful report explains what appears to have changed, what supports that interpretation and what action is justified. “The competitor receives an estimated 40,000 visits” provides very limited context. Showing that its buying guides repeatedly appear for relevant searches, then explaining which purchasing questions they answer, gives the business something more useful and more actionable to consider.
We can demand better collection and clearer disclosure from providers while improving our own interpretation. Understanding the limitations does not let vendors off the hook, and it does not excuse us from checking a conclusion before recommending work.
If your reporting is generating more activity than clear decisions, it might be a good idea to think about strategic SEO consulting. Before buying another tool or changing the whole plan, examine what the existing evidence supports and where the business needs a better answer.
No one should stop using third-party tools and GSC. I would also like us to stop treating a precise-looking number as though it has settled a question it cannot answer. Google’s latest changes are a reason to examine that habit properly. We can do that without declaring the tools dead every time collection gets more difficult.