Keywords Not Provided: What Google Took and How to Work Around It

“Keywords not provided” appears in Google Analytics when organic search data is withheld by the browser or search engine, leaving a blank where keyword-level insight used to sit. Google began encrypting search queries for signed-in users in 2011 and completed the transition to near-total withholding by 2013, which means the vast majority of organic keyword data in your analytics platform is simply gone.

That does not mean you are flying blind. It means you need to be smarter about where you look and what you infer from the data that remains.

I remember the first time I saw “(not provided)” start eating into a client’s keyword report. It was around 2012, and the account manager flagged it as a data quality issue. It was not a data quality issue. It was a policy decision by Google, dressed up as a privacy measure, that happened to also make it significantly harder to evaluate organic search performance without paying for Google’s advertising products. Understanding what happened and what to do about it is still relevant more than a decade later, because the problem has not gone away.

What Does “(Not Provided)” Actually Mean?

When a user searches on Google while signed into their Google account, the search query is stripped from the referral URL before it reaches your analytics platform. Instead of seeing the keyword that brought someone to your site, you see “(not provided)” in its place. Google’s stated rationale was user privacy, specifically protecting signed-in users from having their search queries exposed to third-party website owners.

That is a reasonable position in principle. In practice, the data is still available to advertisers running Google Ads campaigns, which tells you something about the priorities involved. Paid search keywords remained fully visible while organic keyword data disappeared. If the concern were purely about user privacy, that distinction would not exist.

By 2013, Google had extended the encrypted search to all users, not just signed-in ones. The result was that “(not provided)” went from a partial data gap to an almost complete blackout on organic keyword data in tools like Universal Analytics and, subsequently, GA4. Depending on your site and audience, it typically accounts for somewhere between 85% and 100% of organic sessions.

If you are building a broader understanding of how analytics tools work and where their blind spots sit, the Marketing Analytics hub at The Marketing Juice covers GA4, attribution, and measurement strategy in more depth. This article focuses specifically on the keyword data problem and the practical ways to work around it.

Why This Still Matters in a GA4 World

Some marketers have moved on from this problem on the basis that GA4 is a different platform with different capabilities, and that organic keyword data was always imperfect anyway. Both of those things are true and neither of them makes the problem irrelevant.

GA4 does not restore organic keyword visibility. It inherits the same limitation from the underlying data, because the issue is not with the analytics platform, it is with what Google passes through in the referral. If you go to the Traffic Acquisition report in GA4 and filter for organic search, you will see sessions, engagement rates, and conversions, but you will not see which queries drove those results. The keyword column is absent by design.

This matters because keyword-level data is one of the few ways to understand intent at the point of entry. If you know someone arrived via a branded query, that tells you something different than if they arrived via a problem-focused informational query. The conversion behaviour, the content they engage with, the follow-on actions they take, all of it is shaped by what they were looking for when they arrived. Without that context, you are making inferences about audience intent with one hand tied behind your back.

At iProspect, when we were scaling the SEO practice, keyword-level attribution was a constant tension. Clients wanted to know which keywords were driving revenue, not just traffic. We had to build reporting frameworks that triangulated from multiple sources because no single tool gave a clean answer. That discipline of triangulation is still the right approach today.

Google Search Console: The Closest You Will Get to Keyword Data

Google Search Console is the most direct replacement for organic keyword data, and it is free. The Performance report shows you queries, impressions, clicks, click-through rates, and average position over a rolling 16-month window. That is genuinely useful, and it is data that comes directly from Google rather than being estimated by a third party.

The limitation is that Search Console reports at the query level, not at the session or conversion level. You can see that a particular query drove 400 clicks last month, but you cannot see what those 400 people did after they arrived. Did they convert? Did they bounce immediately? Did they read three pages and then leave? Search Console does not tell you that. It tells you what happened before the click, not after it.

This is where the integration between Search Console and GA4 becomes important. You can link the two properties, which surfaces some Search Console data inside GA4’s reporting interface. The Search Console collection in GA4 includes a Queries report and a Google Organic Search Traffic report, which lets you see landing pages alongside query data in the same view. It is not a perfect merge of the two datasets, but it is closer than treating them as entirely separate.

SEMrush has a useful walkthrough of how to surface keyword data within Google Analytics that covers the Search Console integration and some of the workarounds in practical detail. Worth reading if you are setting this up for the first time or auditing an existing configuration.

One thing to be aware of with Search Console: the query data is sampled and filtered. Queries with very low volume are grouped into an “other” category and do not appear in the report. This means the long tail of your organic traffic, which is often where the most commercially specific intent sits, is partially invisible even in Search Console. You see the headline queries, but not the full picture.

Landing Page Analysis: The Practical Workaround

If Search Console gives you the pre-click picture and GA4 gives you the post-click picture, landing page analysis is the bridge between them. The logic is straightforward: if you know which page someone landed on from organic search, you can make a reasonable inference about what they were searching for, because pages that rank well tend to rank for specific intent clusters.

In GA4, go to Reports, then Acquisition, then Traffic Acquisition. Filter the channel grouping to Organic Search and then add a secondary dimension of Landing Page. This gives you a list of pages receiving organic traffic, with engagement metrics attached. It does not tell you the exact query, but combined with what you know about the page’s content and its Search Console query data, you can reconstruct a reasonable picture of intent.

The process is more manual than having keyword data directly in your analytics platform, but it is not guesswork. A product page that receives 2,000 organic sessions per month is almost certainly ranking for commercial-intent queries related to that product. A blog post about a specific problem is almost certainly pulling in informational queries around that problem. The inference is not perfect, but it is directionally reliable.

I have used this approach with clients who were frustrated by the “(not provided)” gap and wanted to understand whether their SEO investment was working. The landing page analysis, combined with Search Console query data for each URL, gave a coherent enough picture to make content and optimisation decisions. It is not the same as having full keyword attribution, but it is enough to act on.

The Crazy Egg guide to building GA4 dashboards has some practical advice on structuring your reporting views to make this kind of landing page analysis easier to run regularly rather than as a one-off investigation.

Third-Party Tools: Useful Estimates, Not Facts

SEMrush, Ahrefs, Moz, and similar tools offer keyword ranking data and organic traffic estimates for any domain, including your own. These tools crawl search results, track positions, and model traffic volumes based on click-through rate curves and search volume data. They are genuinely useful for competitive analysis and for understanding the keyword landscape around your content.

The important caveat is that they are estimates. Their traffic figures are modelled, not measured. Their keyword rankings are snapshots, not continuous. Their data is valuable for directional decisions but should not be treated as ground truth for performance reporting. I have seen clients get into trouble by reporting third-party tool estimates as if they were actual traffic numbers, and the discrepancy between the tool’s estimate and the actual GA4 session count can be significant.

Use these tools for what they are good at: understanding which keywords you rank for, identifying gaps in your coverage, tracking position changes over time, and benchmarking against competitors. Do not use them as a substitute for your own analytics data. They are a different perspective on the same landscape, and having multiple perspectives is useful, as long as you do not confuse one for another.

Moz has a useful comparison of Google Analytics alternatives that also touches on the keyword data problem and how different platforms handle organic search visibility. It is worth reading if you are evaluating whether GA4 plus Search Console is sufficient or whether you need additional tooling.

The Attribution Problem Underneath the Keyword Problem

There is a deeper issue here that the “(not provided)” conversation tends to obscure. Even when keyword data was fully available in analytics platforms, it was not a clean attribution signal. A user might search for a generic term, land on your site, leave, come back via a branded search three days later, and convert. The keyword that gets credit in a last-click model is the branded query, not the generic one that started the experience. The keyword that gets credit in a first-click model is the generic one, regardless of whether the branded touchpoint was what actually triggered the conversion decision.

Neither is right. Both are simplifications. The keyword data that Google withheld was already an imperfect proxy for how organic search actually contributes to business outcomes. That does not mean losing it was fine, but it does mean that the solution to the “(not provided)” problem is not simply recovering keyword data. It is building a measurement approach that is honest about what organic search can and cannot tell you.

Forrester has written thoughtfully about how sales and marketing measurement need to be aligned but not identical, which is relevant here because the temptation is always to push for more granular data rather than better frameworks for interpreting the data you have. The “(not provided)” problem is partly a data problem and partly a framing problem.

When I was judging the Effie Awards, one of the things that separated strong entries from weak ones was the quality of the measurement rationale, not the volume of data cited. Teams that had thought carefully about what their metrics actually measured, and what they could not measure, were far more persuasive than teams that presented a wall of numbers without acknowledging the gaps. The same principle applies here.

What a Sensible Organic Search Measurement Framework Looks Like

Given all of the above, here is how I would structure organic search measurement for a business that wants a clear picture without pretending the data is better than it is.

First, connect Google Search Console to GA4 and make sure the data is flowing correctly. Check that your Search Console property matches your GA4 property in terms of URL format, specifically whether you are using www or non-www, and http or https. Mismatched properties are a common source of data loss that is entirely avoidable.

Second, build a landing page report in GA4 that shows organic sessions by page, alongside engagement rate, conversions, and revenue where applicable. Run this weekly or monthly depending on your traffic volumes. This is your primary performance view for organic search.

Third, use Search Console’s Performance report to understand query-level data for your top landing pages. Export the queries for each URL and map them to the intent categories that matter for your business, broadly: branded, navigational, informational, and commercial. This gives you a rough picture of the intent mix driving your organic traffic.

Fourth, use a rank tracker, whether that is SEMrush, Ahrefs, or something else, to monitor position changes for your target keywords over time. This is a leading indicator. Position changes tend to precede traffic changes, so tracking rankings gives you early warning of organic performance shifts before they show up in your session data.

Fifth, be explicit in your reporting about what you can and cannot measure. If a stakeholder asks which specific keyword drove a conversion, the honest answer is that you cannot tell them with certainty, but you can tell them which landing page the session started on, what the likely intent was, and what the conversion rate for that page is. That is a defensible answer. Pretending the data is more precise than it is will eventually cause problems.

The HubSpot team made a useful point years ago about why marketing analytics and web analytics are not the same thing. Organic keyword data sits at the intersection of both, which is part of why the “(not provided)” problem feels so disorienting. It is a web analytics gap that has marketing analytics consequences.

The Temptation to Stop Asking the Question

The risk with “(not provided)” is not that it makes organic search measurement impossible. It is that it gives people a convenient excuse to stop measuring organic search properly. I have seen this happen in agencies and in-house teams alike. The keyword data is gone, the workarounds are more labour-intensive than pulling a report, and so the organic channel gets measured by a single top-line number, total organic sessions, without any attempt to understand what is driving it or whether it is working.

That is a bad outcome. Organic search is typically one of the highest-value acquisition channels for most businesses, because the traffic has self-selected intent. Someone who searches for a specific problem and finds your content is a different prospect than someone who was served an ad while browsing. Treating that channel as a black box because the keyword data is messy is a commercial mistake.

Early in my career, when the MD said no to the budget I needed, I did not accept that the work was impossible. I found a different way to get it done. The same logic applies here. The data you want is not available in the form you want it. That does not mean you cannot do the work. It means you need to be more deliberate about how you piece together the picture from the sources that are available.

Forrester has written about the dangers of black-box analytics, specifically the tendency to accept opaque outputs without interrogating the underlying methodology. The “(not provided)” problem is a version of that. If you accept the black box, you lose the ability to ask useful questions about your organic performance. The workarounds exist precisely to prevent that from happening.

If you are working through a broader review of your analytics setup, the Marketing Analytics hub covers GA4 configuration, attribution models, and measurement strategy in more depth. The keyword data problem rarely exists in isolation from other measurement gaps, and fixing them together is more efficient than addressing each one separately.

About the Author

Keith Lacy is a marketing strategist and former agency CEO with 20+ years of experience across agency leadership, performance marketing, and commercial strategy. He writes The Marketing Juice to cut through the noise and share what works.

Frequently Asked Questions