Paid search that converts and organic search that does not: the gap is only visible query by query

Two screens, one customer

I ran Google Ads for my own clinic on a small daily budget: a Performance Max campaign and a Search campaign aimed at the high-intent local terms a patient in Bradford types once they have decided they want a treatment. I reviewed the click and conversion data myself, with no agency, and what struck me was that the two reports I needed most lived in different tools, used different words for the same thing, and never appeared on the same screen.

Google Ads told me which searches had triggered an ad, what each click cost and which had led to an enquiry. Search Console told me which searches had shown the website in the unpaid results, and where. Neither could tell me which of the searches I was paying for the website could have earned on its own.

That question can only be answered query by query. This piece sets out what each report contains and hides, why a join on the query text is the only honest comparison, the six situations that fall out of it, and why the result is always an approximation.

What the Google Ads search terms report contains, and what it withholds

A keyword is what the advertiser adds to an ad group; a search term is "a word or set of words a person enters when searching on Google or one of our Search Network sites" ^[1]. The search terms report lists the latter: what was actually typed.

The report is not complete. Google says it includes "search terms that a significant number of people have used, and that resulted in your ad being shown", and that "some search terms that don't have enough query activity are omitted from the search terms report in order to keep with our standards on data privacy" ^[1]. The threshold is not published.

That restriction dates from September 2020, when Google said it was "updating the search terms report to only include terms that were searched by a significant number of users"; until then it had been "quite common to see search terms with one impression or one click" ^[2]. A year later the report was widened to show terms with impressions but no clicks, with the same threshold, and older data below it was to be removed on 1 February 2022 ^[3].

For a local business, the long tail of searches with a street name or an odd spelling is the part most likely to be missing. The clicks were charged. The words were not shown.

Performance Max reports search terms differently

Performance Max has its own search terms report, with data "available starting from March 2023", and Google advises advertisers to "prioritise conversion data over clicks and impressions" when judging it ^[4]. It does not use keywords; search themes "provide a broader signal to Google's AI" ^[5]. When a Search campaign and a Performance Max campaign could both serve the same query, Google's stated order is that an exact match keyword identical to the search term wins, then identical phrase and broad match keywords or search themes, then AI relevance, then Ad Rank ^[6]. A business running both, as I did, will find the same query recorded under either.

Brand searches can be kept out of both. Brand exclusions "automatically block your ads from showing on searches for your brand name, common misspellings, and even related subsidiary brands", to "avoid paying for clicks from users who are already navigating to your site" ^[7].

What Search Console contains, and the near half it does not show

Search Console's Performance report counts clicks as "the number of times a user clicked your site from Google Search results", impressions as "how many times your site appeared in Search results", and average position as "the average position of the topmost result from your site" ^[9].

The larger gap is deliberate. Google states that "some queries are omitted from the report to protect user privacy. These are called anonymized queries", and that they are "included in chart totals, unless a query filter is applied" ^[10]. The total at the top of the screen is right; the rows beneath do not add up to it, and the missing rows are the rare searches.

How much is missing has been measured. Ahrefs examined 22 billion clicks across 887,534 Search Console properties and found anonymised queries made up 46.77 per cent of traffic in April 2025 ^[11]. The pattern by size is the uncomfortable part for a small firm: "sites with lower traffic and sites with higher traffic seem to be missing more data", and the most frequent share for a single site "falls between 45% and 80%" ^[11].

So both sides hide the same kind of query, for the same stated reason. Any join of the two covers only the searches common enough to survive both filters.

Google's own paid and organic report, and where it stops

Google does provide a report that puts the two side by side. The paid and organic report in Google Ads, available once Search Console is linked, shows "how often pages from your website are showing in Google's free organic search results, and which search terms triggered those results" ^[8]. Its published columns are organic clicks, organic queries, ad clicks, ad impressions and a combined count, and Google suggests using it to "discover potential keywords" and to "observe the impact across text ads, organic results, and combined traffic" ^[8].

It is a useful table and deserves more use than it gets. But the columns Google lists are about clicks and impressions. They do not, in that list, set cost and conversions per query against organic position, and that is the comparison a budget decision needs. The report also inherits the omissions of both its sources. It is a good place to start and a poor place to stop.

Why the honest comparison is a join on the query itself

The alternative is to take both exports and join them on the words. Take every Google Ads search term with a handful of clicks in the window, and every Search Console query with impressions in the same window. Normalise both: lower-case, strip punctuation, collapse plurals and "near me" variants, and remove the business's own brand names so that people who already know the clinic are not counted as market demand. Join on the normalised query. Each row then has a paid side (clicks, cost, conversions) and an organic side (impressions, clicks, position), and either may be empty.

An empty side is not a failure of the join. It is the finding: a query with paid conversions and no organic row is a subject the website has never been shown for, and a query with organic impressions and no paid row is demand no campaign has been pointed at.

This is what the HighRegard application does, and it is worth describing plainly. The demo workspace runs the join on a synthetic clinic, with "one row per query intent, combining paid search terms with organic Search Console rows" over a 28-day window, and shows paid clicks, spend, conversions, organic clicks, impressions, position and a scenario label for each row ^[20]. The page states that "every number here is synthetic and generated to exercise the analysis" ^[20]. The labels are the six situations below. Where an account's conversion tracking looks unreliable, the application withholds the pause test described under the fourth situation, because it cannot be read without a trustworthy conversion count.

The six situations that fall out of the join

The table sorts every query into one of six situations.

1. Converting in paid, invisible organically

The query converts in paid search and the site ranks outside the top ten organically, or not at all. This is the row that pays for the exercise: proof of commercial intent, bought with real money, set against a page that is not competing for it. The demo's clearest example is "laser hair removal near me": 105 paid clicks for £141.34 and 10 conversions, alongside 11 organic clicks from 1,101 impressions at an average position of 14.4 ^[20]. In Google Ads alone the campaign looks fine. In Search Console alone a mid-teens position with a dozen clicks is unremarkable. Only the join says the business is paying for demand it could try to earn.

For content, build or strengthen the page for that intent. For spend, keep paying until organic clicks arrive, then test reducing the bid. Do not switch the ads off on the day the page goes live; nobody knows on that day whether it will be shown.

2. Paid only

Converts in paid, no organic impressions at all. Often the site has never answered the query in words: a price, a specific procedure, a town it serves but does not name. The fix is usually a page, not a bid. It is the largest group in the synthetic account, at 26 queries ^[20].

3. Organic demand, no paid

Impressions and a reasonable position, but no campaign targets the term. The demo calls it "a candidate for a controlled paid test if the page converts" ^[20]. If it does not convert, more visitors will not explain why.

4. Organic already winning where the business pays

The site already ranks in the top three and the campaign is still spending on the same term. This is the cannibalisation case, where the published evidence is richest and least settled. The honest test is a controlled one: alternate weeks with the term paused and compare total conversions, paid plus organic. If the paused weeks lose little, the bid can come down.

5. Expensive clicks with organic within reach

High cost per click on a query where the site ranks between roughly four and fifteen. Small organic gains here can replace disproportionate spend; it is a content priority before it is a bidding one.

6. Balanced

Both channels present, neither obviously wasteful. Leave it alone and come back next month.

Brand terms sit outside the six because they were removed before the join, for reasons discussed below.

What Google's own research says about paid and organic clicks

Google has published its own answer to the cannibalisation question, and it deserves careful reading rather than repetition as a slogan. In 2011 Google's quantitative team reported that across more than 400 "search ads pause studies", in which organic clicks were observed while ads were off, "a full 89% of the traffic generated by search ads is not replaced by organic clicks when ads are paused" ^[12]. A 2012 update of more than 5,300 cases put the figure at 85 per cent where ads were paused, and noted monthly volatility driven by "the mix of advertisers who choose to pause their ads in that month" ^[13].

The most useful paper for the six situations split the result by organic rank. Across 390 studies between April and October 2011, Google found that "81% of ad impressions and 66% of ad clicks occur in the absence of an associated organic search result", and that the incremental share of ad clicks was 50 per cent (plus or minus four points) with a rank one organic result, 82 per cent (plus or minus 14) at ranks two to four, and 96 per cent (plus or minus 11) at rank five or lower ^[14].

On average, 50% of the ad clicks that occur with a top rank organic result are incremental ^[14].

That is a strong argument for situations one and two, where the organic result is absent or distant, and a much weaker one for situation four, where a top organic result is present and around half the paid clicks would have arrived anyway ^[14].

The other side deserves stating. The research is Google's own, on Google's advertisers, from 2011 and 2012, when the results page looked very different from the one a Bradford patient sees on a phone today. The figures are averages, and the authors themselves flag volatility in the sample ^[13]. A local business can only run its own pause test.

What independent research says, especially about brand terms

Independent studies have been less kind, to brand bidding in particular. A field experiment at eBay by Blake, Nosko and Tadelis, circulated as an NBER working paper in May 2014, found that "brand-keyword ads have no measurable short-term benefits" and that for non-brand keywords "new and infrequent users are positively influenced by ads" while frequent users "account for most of the advertising expenses, resulting in average returns that are negative" ^[15].

A later study by Simonov, Nosko and Rao, published in Marketing Science in 2018, found the brand picture depends on who else is bidding: with no competitors present, brand ads had "a positive, statistically significant impact" of one to four per cent, but where the brand did not advertise while competitors did, competitors took 18 to 42 per cent of clicks ^[16].

The eBay finding is situation four at scale: a household name at the top of the organic results, paying for clicks it would have received. The Simonov result is why brand terms are decided separately: the answer depends on whether a competitor is bidding on the clinic's name, which the table cannot see but a single search can.

On the overlap that decides which report a paid query lands in, Adalysis examined over 3,300 non-retail Performance Max campaigns and reported that 45 per cent of their search terms could show ads from either campaign type ^[17]. That is one firm's client data, not a controlled study, but it is a reason to pull both search terms reports into the join.

Where the join is lossy

The method has three honest limits.

First, both sides hide low-volume queries. Google Ads omits terms without "enough query activity" ^[1], Search Console omits anonymised queries that are still "included in chart totals" ^[10], and the independent estimate is that close to half of organic traffic sits behind that veil ^[11]. The rarest local phrasings, often the most valuable, are the most likely to be absent from both.

Second, the two sides count differently. Google Ads "attributes these conversions to the date of the ad click" ^[18], so a paid row's conversions belong to the day someone clicked, not the day they enquired. Google's own explanation of why ad clicks and analytics sessions disagree lists repeated clicks, pages that never finished loading, and invalid clicks that Google Ads filters out but analytics records ^[19]. Search Console counts clicks on the results page and knows nothing of what happened afterwards ^[9]. A paid conversion and an organic click are not the same unit.

Third, a click is not an enquiry, and an enquiry is not a patient. I learned this on my own account, when a broken conversion tag made a working campaign look like a failing one. Nothing about the search terms, the bids or the landing page had changed; the tag had. A join that reads conversions from Google Ads is only as honest as the tag feeding it.

What this means for a local business

For a clinic, a solicitor or a plumber with a modest budget, the sequence is short.

1. Link Search Console to Google Ads and open the paid and organic report, which shows organic and ad clicks against the same search terms ^[8].
2. Export the search terms report from every Search and Performance Max campaign, because Google's priority rules decide which campaign a query is recorded under ^[6].
3. Export Search Console queries for the same window, remembering that anonymised queries are counted in the total but not listed ^[10].
4. Normalise both, strip the brand, join on the query, and label each row with one of the six situations.
5. Act on situations one and two first: they are proof of intent with no page to serve it.
6. Treat situation four as a hypothesis to test with a pause, and only once conversion tracking has been checked.
7. Decide brand terms separately, after searching the business name and seeing who else appears.

None of this promises a position, a place in the local pack or a number of enquiries. What it offers is smaller: spending and content decisions made from the same list of words, in the same window, with the omissions written down.

What the evidence does not settle

The incrementality figures most people quote are Google's, more than a decade old, and reported as averages with wide margins at the ranks that matter most ^[14]. No published study I could open reports the same split for small local advertisers on a modern mobile results page with map results and AI features above the organic listings. The direction of the effect is plausible; its size for any one business is unknown.

The independent brand-bidding evidence points in two directions depending on competition ^{[15] [16]}, and neither study concerned a local service business whose competitors may or may not be bidding on its name in a given month.

The share of queries hidden on the organic side is measured in aggregate ^[11] but not for any single site, and the paid-side threshold is not published at all ^[1], so the join's coverage can only be bounded.

Finally, none of this settles what a conversion is worth. Ten conversions for £141.34 of spend mean nothing until someone has said what a new patient, client or job is worth over time, and that figure is in neither report ^[20].

What remains uncertain

When I looked at my own clinic's paid and organic data together, the numbers did not tell me what to do. They told me which questions I had been unable to ask. That is a smaller claim than most dashboards make, and it is the one I am prepared to make.

The join is lossy on both sides, the units differ, the best-known incrementality figures are Google's own and old, and the answer for brand terms depends on competitors the table cannot see. What survives is still the most useful thing a small advertiser can learn from an ads account: the searches people paid to be found for, set against the searches the website was shown for, and the gap between them. How much of that gap a page can close rather than a bid, each business will only discover by trying, one query at a time.