The biggest name in the category buys the least search

Live US search text ads, four advertisers we have taken apart
AdvertiserCategoryText ads
monday.comB2B software~5,000
GymsharkDTC apparel~400
HeadspaceSubscription app~400
RunnaSubscription app~200
DuolingoSubscription app66

Source: Google Ads Transparency Center, US region, text format. Each checked on its own teardown date; Duolingo checked 20 September 2026.

Runna is a running-plan app owned by Strava. Duolingo is a listed company with hundreds of millions of downloads. On this one surface, Runna buys three times as much.

The tempting explanation is that Duolingo does not need to, because everyone already knows it. That is half right and the useful half is the other one: paid search does not create demand, it intercepts demand that already exists in words. Almost nobody sits down and searches for a language app with a wallet out. They download one on a whim, usually from the store or because a friend mentioned it. There is no commercial query to buy.

Which is why the money goes where there is one.

Half the search advertising is for a different product

The Duolingo English Test is a paid proficiency exam, sold from englishtest.duolingo.com, and it has its own ads with its own sitelinks – Take a Practice Test, Free Practice Tests For You, Share Results Unlimited Times, How Scores Work, Who Accepts the Test.

Its description does something the language app never does. It names a price and compares it:

Testing made easy, no appointment needed. Take the test online, anytime, anywhere. The DET is only $70 – most other tests cost over $200. Save your money for study abroad.

Read what that ad assumes about its reader. They know they need an English test. They know the alternatives. They have a deadline, because a university application has one. They are comparing on price and inconvenience, and the ad answers both in a single line – no appointment, a third of the cost.

That is a commercial query with a deadline attached, and it is worth buying. "Language app" is not.

02

What the app ads actually say

Two shapes, and both of them are about money rather than about learning.

Observed text ads, US
HeadlineDescriptionWhat it is selling
Try 1 Month For $0 – Start Your Free Month$0 for 1 Month of Super Duolingo. Learn without ads and practice without limits.The paid tier, priced at zero for a month
Speak English – Free Language LearningThe Free, Fun, And Effective Way To Learn A Language. Duolingo – the world's most popular…The free app, to someone searching a language rather than a brand
Duolingo English Test – Test Online, Anytime, AnywhereThe DET is only $70 – most other tests cost over $200. Save your money for study abroad.A $70 exam, with the competitor price in the copy

Source: Google Ads Transparency Center, advertiser Duolingo Inc, verified. Checked 20 September 2026.

The first one is the interesting one for anyone running a subscription app. “$0 for 1 month” is not a description of a free trial, it is a price. It reframes the upgrade as a transaction that happens to cost nothing this month, which sets the expectation that it will cost something next month – and that expectation is most of what decides whether a trial converts.

Compare it with the trial language in our Runna teardown, where the ad leads with "first week free" and "no credit card required". Both put the trial in the creative. One frames it as a gift, the other as a price. The first buys more trial starts; the second buys fewer and better ones. Neither is correct in the abstract – it depends entirely on whether your product can prove itself before the charge lands, which you can work out with the payback calculator rather than by arguing.

The second ad, "Speak English – Free Language Learning", is aimed at someone who typed a language rather than a brand. It is the only ad in the set doing category acquisition, and there are very few of them, which is consistent with everything above.

03

What we would ask if this were our account

None of it is answerable from outside, and all of it is answerable in a fortnight from inside.

  1. How much of the 66 is brand defence? On a name this famous, a large share of search volume is people typing the brand. That budget is a toll rather than acquisition, and it should be reported separately or it flatters everything else.
  2. Does the English Test traffic pay back faster than app traffic? Almost certainly, since one is a $70 purchase with a deadline and the other is a free download that might subscribe later. If so, the split is not 50/50 for long.
  3. What does “$0 for 1 month” do to month-two retention against a card-gated trial? The trial-start number will look better. The cohort that matters is the one still paying in month three.
  4. Is there a second product hiding in the catalogue with the same shape as the English Test? Something paid, compared, and bought under a deadline. That is where the next search budget goes, and it is a question every subscription business should ask once a year.

Questions people actually ask

Sixty-six ads – is the library missing something?
It is showing US text ads for duolingo.com and its subdomains, which is what we asked it for. App install campaigns on other networks, Apple Search Ads and anything outside the US are not in this count, and none of them are public. What the figure does tell you is how little of the search surface the company chooses to buy.
Why does a free app advertise at all?
Because the free app is not the product being sold in these ads. Half of the observed search advertising is for the Duolingo English Test, which costs money and is bought by people with a deadline. The free app appears mostly as a $0 trial of the paid tier.
Does this mean brand-led companies should skip paid search?
No. It means paid search follows demand that already exists in words. If people are searching for the problem you solve, buy it. If your category has no search volume with a wallet behind it, no amount of budget invents any, and the money belongs somewhere else.
Is a price comparison in an ad a good idea?
It is when your price is the argument and the reader already knows the alternatives cost more. It is a bad idea when it starts a race you cannot win, or when the reader has no reference point and the number just looks expensive out of context.
Do you know whether any of it works?
No. Spend, conversion rates and subscriber numbers are private. This reads decisions rather than outcomes, and where we infer a reason we say we are inferring.

This, on your account, with the numbers attached

The free audit is this analysis run on your own advertising – except we can see spend, trial conversion and the cohort data underneath, rather than inferring from what the libraries publish.