Not Provided Keywords: What They Mean and How to Recover the Data

“(not provided)” keywords are organic search visits whose search query was withheld by Google, so your analytics knows the visitor came from Google but not what they typed. It is a privacy feature, not a bug, and nothing in your analytics setup will reverse it. The keyword data still exists in aggregate in Google Search Console, and the practical fix is to join those queries to your conversions through the landing page.

This guide explains why it happens, what it looks like in GA4 and other tools, every place you can still get query data (and what each one hides), and a worked example that turns Search Console queries into an estimated value per keyword.

What “(not provided)” keywords mean

When someone clicks a search result, their browser tells your site where the click came from through the referrer. Years ago, that referrer carried the full search URL, query included, so analytics tools could print a list of keywords next to every visit, conversion and sale.

Today Google sends only its origin (for example, https://www.google.com/) with no query. Your analytics can still classify the visit as organic search from Google, but the keyword field is empty. Tools fill that empty field with a placeholder: Universal Analytics made “(not provided)” famous, and you will see the same idea as “(not provided)”, “(not set)” or simply a missing keyword column in other tools.

Why Google hides the keyword

It started with encryption. On October 18, 2011, Google announced it would move signed-in users to encrypted search on https, and said plainly that sites receiving clicks from organic listings would “still know that you came from Google” but would not receive the individual query (Making search more secure, Official Google Blog). The same post pointed site owners to an aggregated list of their top 1,000 queries in what was then Webmaster Tools, the ancestor of Search Console.

Encryption later became the default for essentially all Google searches, so the share of organic visits with a visible keyword shrank to almost nothing. Two details from that announcement still shape your reports:

• Ads were treated differently. The post said ad clicks would keep sending the query so advertisers could measure campaigns. That is why paid search reports still have search terms and organic reports do not.
• The replacement was aggregate data. Query data moved from per-visit (in your analytics) to per-site totals (in Google’s own tool). Everything below follows from that split.

What (not provided) looks like in your reports

The label differs by tool, but the pattern is the same: organic search sessions with no query attached.

If you see (not provided) on traffic that is not organic search, the cause is different. In GA4, missing campaign parameters or a missing session start event produce (not set), which is fixable. That is a tagging problem, not the keyword privacy problem described here.

How to recover not provided keywords: six sources

None of these gives back the query for an individual visit. Together they give you a good picture of which queries bring traffic to which pages.

1. Google Search Console Performance report

This is the primary source. It lists queries with clicks, impressions, click-through rate and average position. The useful move is to filter by a single page first, then open the Queries tab: you get the queries that lead to that page, which is what you need to connect keywords to outcomes later.

2. GA4 linked to Search Console

Linking adds two reports to GA4: Google Organic Search Queries and Google Organic Search Traffic. Read the fine print, though. Google’s documentation says Search Console metrics are only compatible with Search Console dimensions plus three Analytics dimensions: Landing page, Device and Country, and the queries report cannot be broken down by Analytics dimensions at all (Connect Search Console to Google Analytics, Analytics Help). The data is also limited to the 16 months Search Console keeps and arrives about 48 hours late.

In practice: linking puts queries and conversions in the same tool, but not in the same row. The landing page is the only bridge.

3. Bing Webmaster Tools

Bing has its own search performance report with keywords, clicks and impressions for your site. Bing’s volume is smaller than Google’s for most sites, but its query list is a free second opinion on how people find you.

4. Your own site search

What visitors type into your search box after they land is first-party data you fully control. GA4’s enhanced measurement can record site search terms. These are not the Google queries, but they show the words your audience uses and what they could not find on the landing page.

5. Google Ads search terms (if you run ads)

The search terms report shows the queries that triggered your ads, with conversions attached. It is the only Google source that ties queries to conversions directly. It also feeds your negative keyword list. Treat it as a hint about organic demand, not a copy of it: ads match differently and you only see the queries you paid for.

6. Third-party keyword and rank tools

Ahrefs, Semrush, Moz and similar tools estimate which keywords your pages rank for, using their own crawls and clickstream models. They are good for discovering keywords you could rank for and for competitor research. They are estimates: they will miss many long-tail queries you actually get, and their traffic numbers will not match Search Console.

Which source answers which question

Search Console’s own gaps, and how to debug them

Search Console is the best answer to (not provided), but it does not show everything. Google’s Search Central team spelled out the limits in October 2022 (A deep dive into Search Console performance data filtering and limits):

• Anonymized queries. Queries not issued by more than a few dozen users over two to three months are left out of the tables to protect privacy. They still count in the chart totals.
• Filters drop them. As soon as you filter by query, anonymized queries are excluded, so “contains X” plus “does not contain X” will add up to less than the unfiltered total.
• Row limits. The interface exports at most 1,000 rows. The Search Analytics API (and the Looker Studio connector) goes up to 50,000 rows per day per site per search type.

Turn (not provided) into revenue by landing page

The question behind most (not provided) searches is not really “what did people type”. It is “which searches make us money”. Keyword-level revenue for organic traffic is gone, but page-level revenue is not, and every query in Search Console belongs to a page. So measure conversions and their value per organic landing page, then use the page’s query mix to describe what that revenue is made of. The pillar guide on SEO conversion tracking covers the setup, and how to calculate conversion value covers putting a dollar figure on leads.

How to estimate keyword value, step by step

1. Pick the landing page and date range. One organic landing page, the same dates in Search Console and in your analytics.
2. Get the organic conversion value for that page. Conversions and their value from organic sessions that started on that page.
3. Export the page’s queries from Search Console. Filter by the exact page URL, export queries with clicks, and note the page’s total clicks from the chart (the unfiltered-by-query total).
4. Compute value per click. Divide the page’s organic conversion value by its total Search Console clicks.
5. Allocate value to queries. Multiply each query’s clicks by value per click. Keep the unlisted remainder as its own row instead of spreading it over the visible queries.

Worked example (illustrative numbers)

A software company’s comparison page received organic visits that produced 30 demo requests in a month. Each demo request is valued at $150, so the page’s organic conversion value is 30 × $150 = $4,500. Search Console shows 1,100 total clicks to the page that month, with these queries itemized:

Value per click is $4,500 ÷ 1,100 = about $4.09. The top query is worth roughly 400 × $4.09 = $1,636. Note the last row: 300 of the 1,100 clicks (27%) came from queries Search Console does not itemize. Keeping that row visible stops you from overstating the value of the queries you can see.

Two adjustments make the estimate more honest:

• Split branded from non-branded. Branded queries usually convert better than generic ones, so an even split flatters generic terms. If a page gets both, separate them first; the guide on branded vs non-branded traffic shows how.
• Prefer pages with one dominant intent. The method works best when a page ranks for a tight cluster of similar queries. A page that ranks for unrelated topics gives a muddy average.

SEOConversion is built around this page-level view: it reports conversions and their value from organic search and AI assistants by landing page, because search engines hide the keyword. It does not show keyword-level conversions, and no tool honestly can for organic traffic.

AI assistants repeat the (not provided) problem

Visits from ChatGPT, Perplexity, Gemini and other assistants carry no prompt at all. At best you get a referrer or a UTM tag that names the assistant; at worst the visit has no referrer and lands in Direct. The approach is the same as with (not provided): measure by assistant and landing page, not by prompt. See utm_source=chatgpt.com for what ChatGPT passes, and what direct traffic means for why some of these visits look like Direct.

Myths about not provided keywords

• “It is a broken tracking setup.” No. Google withholds the query on purpose. A correct setup still shows it.
• “Another analytics tool will bring keywords back.” No analytics tool receives the query, so none can report it per visit. Tools that show organic keywords are importing Search Console data or estimating.
• “Search Console shows every query.” It omits anonymized queries and caps exports, as covered above.
• “Rank tracker keywords are my real keywords.” They are the keywords the tool tracks or models. Your real long tail is mostly invisible to it.
• “Filling in meta keywords helps.” Google ignores the meta keywords tag. It changes neither rankings nor reporting.

For the attribution side of the same problem, which organic visits get credit for a sale and when, read organic conversion attribution.

FAQ