The High Cost of Hidden Search Terms: How Google Skims Profit from Advertisers

Google claims that search terms are hidden for privacy reasons and not performance reasons, so we wanted to find out: Do hidden search terms perform the same as visible ones?

After analyzing $20,000,000 ($19,626,409 to be exact) in search and shopping spend, we found that for every dollar that you spend on Google Ads, they skim almost a dollar in forced inefficiency through hidden search terms. If you spent $1 million on ads and generated $4.75 million in revenue, the blended ROAS in our dataset is 4.75x, your ads would have generated over $5.6 million in revenue if not for the performance delta between visible and hidden search terms.

The fact that Google is skimming for their own benefit isn’t shocking, but the amount they are skimming is. We know that Google manipulates their platform to their benefit at the expense of advertisers. Exhibits from the anti-trust trial prove this.

Another way that Google scrapes profit from advertisers is through the heavy use of hidden search terms. Until September 2020, Google shared almost all search query data. In 2020, Google decided that it would reduce the percentage of search query data that advertisers were allowed to see for “privacy reasons.” I simply do not believe that 28% of all searches coming into Google Ads need to be hidden for privacy reasons.

Since that change, advertisers have seen the percentage of visible search queries plummet. In our data set we found that 26.7% of search spend and 33.5% of shopping spend is now hidden.

Visibility varies widely. Among the accounts in our MCC, the account with the lowest hidden percentage had only 12.6% of their keyword volume hidden. The highest had 73.3% hidden.

Getting all of this data was a herculean task. There is no easy way to get this data with systems, so we had to pull search query report data manually. We pulled this data for 933 campaigns: 779 search campaigns and 154 shopping campaigns. We then categorized and analyzed it so we could mine it for insights.

There are an infinite number of ways we could look at cross sections of data, so we tried to keep it as broadly applicable as possible. If there’s specific data you want to know about that you don’t see in the piece, send me an email or hit me up on Twitter/LinkedIn, and I’ll try to get it for you.

I hope you find this information enlightening.

Expected Vs Unexpected Headline Numbers

I came into this research project with the expectation that hidden search terms would be worse than visible search terms. The data bears that out. However, there were some numbers that surprised me. Before I dive deeper into the data and our findings, I wanted to highlight some of these numbers.

The Expected

We expected to see visible outperform hidden as this is something that, anecdotally, everyone who works with Google ads has seen. The size of the delta between the two, however, (5.60x for visible and 2.59x for hidden) was even larger than I was expecting. For all data in our dataset, visible search terms perform over two times better than hidden search terms. This means that 28% of spend on our sample is performing half as well as the other 72%.

Likewise, we expected to see CPCs be higher for hidden search terms, and that has also been borne out in the data. The main benefit to Google in hiding search terms is that it severely restricts the ability of advertisers to opt out of the auction. This increases the number of advertisers competing in the auction for these search terms, even if they don’t want to do so, and increases the CPCs. As a result, we find that Visible CPCs are $1.24 and hidden CPCs are $1.88.

What’s hiding in those elevated CPCs is increased profit for Google by monetizing search queries that would not be receiving this level of auction competition otherwise.

CTRs being higher for visible search terms should come as no surprise either. It makes sense that search terms that are visible are more relevant, in general, than search terms that are hidden. If they were highly relevant to the query, they wouldn’t be hidden in the first place. The difference between the two is stark with visible search terms earning a CTR of 3.78% on average and hidden terms earning just a 2.13% CTR.

A lower CTR and a higher CPC tells the tale of low quality terms with forced bidding. This is based on almost 3,000,000 hidden clicks too, so the sample size is quite large.

Overall, the aggregated data was in line with what I know most of us experience. Hidden search terms being worse from a performance standpoint than visible search terms.

With such a large dataset, we were also able to dig deeper into the data and still have statistically significant results, including some unexpected results.

The Unexpected

Without a doubt, the most unexpected data that came out of this project was the fact that for nonbrand search and shopping, the average ROAS between visible and hidden terms was incredibly close. The average visible ROAS for nonbrand was 1.90x and 1.88x for hidden.

This was driven in large part by nonbrand shopping ads (22.7% of all nonbrand spend went to shopping) which actually saw a higher ROAS for hidden search terms than visible search terms at 2.64 vs 2.51.

Key Findings At A Glance

Across all metrics and client types, visible queries consistently outperform hidden ones. ROAS is 116% higher, conversion rates increase by 49.4%, CTRs improve by 77.5%, and CPCs decrease by 34%.

The trend on all this data is clear: Google is hiding lower quality search terms that advertisers would be less likely to bid on if they were visible.

Performance by Client Type

We segmented our conversion actions into two different types: purchases and leads. Most of our clients are e-commerce, with 75.54% of the spend in our data set being for purchase conversions.

For the following data, we used only this purchase conversion data so that it would be a fair comparison between search and shopping.

Breaking It Down: Brand vs Nonbrand, Shopping vs Search

As you’ll see in the data below, the trend is that the delta between hidden and visible search terms is greater with Search than with Shopping. The delta is larger with Brand than it is with Nonbrand. Therefore, and as expected to those in Google Ads accounts often, most brazen and atrocious forced inefficiencies are to be found in Branded Search campaigns.

All brand

Brand spend accounted for 33.63% of the total spend across our dataset. Of that spend, hidden terms had a ROAS 57% lower, a conversion rate 35% lower, a CTR 61% lower, and a CPC 99% higher, it’s worth wondering how much brand is actually in the hidden brand terms. Match types are nearly meaningless with even exact match frequently matching to hundreds of different queries, so it feels like there’s nonbrand queries in the hidden search terms on brand campaigns. The hidden brand CPC of $1.53 is closer to the hidden nonbrand CPC of $1.83 than it is to the visible brand CPC!

This sample is very large too. Hidden search terms spent $1.3M which represented a forced revenue inefficiency of $8.7M.

All nonbrand

The closeness of hidden vs visible terms for nonbrand was the most surprising information that we found in the dataset. It feels like every time you look at a nonbrand campaign hidden and visible are starkly different. I wondered if we had outliers in our data, but looking at the median it is still 1.24x for visible and 1.17x for hidden.

The easiest answer is that Google does a great job of understanding nonbrand intent and is able to effectively monetize queries regardless of whether they show you the query or not.

There is also the possibility that misspelled brand terms are leaking in as well which could be polluting the data. It’s impossible to know because all that data is hidden for “privacy reasons.”

All shopping

The most interesting data point for me for shopping performance is the fact that we see a higher CTR and a lower CPC. These two rules held true for both brand and nonbrand breakdowns.

I’m really not sure about why this has been happening around shopping campaigns.

Brand shopping

Brand shopping performance is much closer than brand search. We looked at $1.35M in spend against brand shopping overall. The median was 4.3x for hidden vs 5.1x for visible, so pretty in line with our averages.

CTR being 16.13% higher on hidden search terms stood out to me when looking at this data.

Nonbrand shopping

The most surprising piece of data that we uncovered in our research is the fact that nonbrand shopping campaigns had a better ROAS for hidden search terms (2.64x) than visible search terms (2.51x).

With higher CTRs and lower CPCs, it makes me wonder what kinds of queries are making their way into the hidden search terms. Is there some percentage of misspelled brand terms that are not being excluded by the negative keyword lists? There’s no way to know for certain. I’m not sure what other hypotheses would generate this result. Being much more aggressive about brand negatives in terms of misspellings and groupings is something we plan to test as a result of these findings.

Search All

The overall picture on search is stark, though this is largely driven by the delta on brand search. ROAS is a whopping 151% higher for visible queries than hidden ones, with conversion rates up 69%, CTRs up 126% and CPCs 43% lower.

Search brand

The bulk of this delta lives at the brand level. Brand search should be absolutely cooking, and it is when those search terms are visible.

Search nonbrand

While not as surprising as the shopping data, the performance of hidden terms on nonbrand search was much better than I would have expected. With higher CPCs and lower conversion rates, the performance is as close as it is because of a much higher average conversion value.

My Takeaways on the Data

This is clearly another step in Google’s march to become a flat tax on businesses. Almost every update that comes out removes more control and forces inefficiencies, one way or another, that are always detrimental to advertisers but beneficial to Google. We know this is true because Google is a publicly traded company, and they have a responsibility to make money for shareholders, and much of that increase in revenue comes from advertisers paying more for the same stuff. It’s not surprising that Google is doing this but the degree to which Google is taking money out of advertisers’ pockets is.

This picture also undersells the issue of low quality inventory getting pushed by Google. This excludes all Performance Max spend as Performance Max has not given this level of visibility until very recently. It strains credulity to believe that Performance Max campaigns are more favorable to advertisers when it comes to spend against low quality inventory than search or shopping. We have found through numerous rounds of holdout testing that despite generally higher in-platform performance than search and shopping, incremental performance on Performance Max lags way behind. Therefore, if you are running Performance Max campaigns, it is a safe assumption that your efficiency loss to waste is greater than what our dataset shows.

The colossal delta on brand is definitely surprising. It really makes me wonder what even exists in the hidden terms in the first place. It would be my supposition that there is a decent amount of nonbrand that is leaking in there because of the way that match types work. We find that even with exact match there can be hundreds of “close variant” search terms that find their way in. Brand volume is also not insignificant. According to research done by Sparktoro, over 44% of all Google searches are for brands.

While there were definitely surprises in terms of performance of hidden and visible terms being close together in some structures, namely nonbrand shopping, I believe that the evidence points to Google hiding search terms not for privacy reasons but in an effort to increase revenue.

Arguments In Favor of Hidden Data

Some will argue that our analysis is unfair. That there are arguments in favor of the hidden search queries. Google will argue that it is for privacy reasons. That advertisers can’t see almost 30% of the ad volume that they pay for because of “privacy”. And I’ve graced the cover of People’s Sexiest Man Alive issue.

The other main argument that Google could make about the hidden data is that it is net additional volume that advertisers wouldn’t otherwise be getting. That if it performs above target, who cares if it is hidden.

The response to this is, “If it’s quality traffic and performs well, it should be visible to advertisers to make the decision for themselves about the quality.”

Advertisers aren’t stupid. They don’t need search query data hidden from them because they won’t understand what is good and what is not. Access to this data was available for over a decade, and it never caused any issues.

The real answer is likely that Google wanted to increase the amount of bidders for low quality search queries and by hiding the queries they were able to both reduce the number of advertisers excluding those queries and increase the quantity of poorly monetizable queries advertisers were bidding on without being able to see it.

This aligns with the data. Hidden clicks are 51.6% more expensive on average than visible clicks. Their CTRs are 43.7% lower. Higher CPCs with lower CTRs points to queries that are less relevant with higher competition. Exactly the sort of behavior you would expect to see if Google were using hidden terms as a way to push advertisers into auctions for terms that are less relevant.

Where to go from here

Unfortunately, there is not a lot of recourse for advertisers. Google is considered to be operating as an illegal monopoly by the US justice system and as such, cannot be influenced by market dynamics. There isn’t anywhere else to take your ad dollars.

The most important steps that advertisers can take are to not trust Google to have their best interests at heart and to avoid utilizing products that further obfuscate performance unless the performance is verified by third party measurement as being more effective. Products like Performance Max. The more data Google can hide from you, the more they can skim for their own benefit.

Google is trying to squeeze every last dollar out of advertisers because they already own the entire market. In Q1 2025, “Search & Other” accounted for $50.7 billion in revenue, approximately 56% of total revenue for Google in Q1 which is a 10% increase YoY. They are generating these results by maximizing how much advertisers will spend on their entire ad inventory. This means the advertiser’s relationship with Google is inherently adversarial. I wrote about the details of the relationship here.

It pains me to be so critical of Google as someone who has built a 15 year career running ads on the platform, but the numbers don’t lie. Fellow old timers in the industry have known for quite some time that this is the way Google has been going, I just had the 14,000,000 data points to prove it.