Menu
Google Ads audit case studies

What Claude found in 10 real Google Ads accounts.

Ten advertisers in 6 countries ran Boxoo’s free audit, from a personal trainer spending 113 BRL a month to a clinic running Demand Gen. Claude filed 231 cases. Here is what came up, account by account, with the numbers.

Every check is done by ClaudeAudit my account free Explore the demo
The Boxoo home screen after an audit: cases ranked by priority, with a summary of the account
  • Real accounts, anonymised
  • Numbers exactly as Claude reported them
  • The audit is free, no card
  • Read-only until you press Apply

The audits in numbers

Every case is a finding with evidence from the account: what is wrong, how much it costs, and the fix. Here is where they landed.

10real accounts auditedin 6 countries
231cases filedeach with the numbers behind it
37high-priority casesproblems costing money now
7 of 10accounts had a tracking problemthe most common area
Cases by area across 10 real audits
  • Cases filed
  • Most common area
Conversion tracking39
Ads and assets38
Search terms27
Keywords26
Growth opportunities24
Recent changes18
Budgets and reach16
Google’s recommendations14
Structure and settings8
Landing pages and speed6
View as table
ItemCasesGroup
Conversion tracking39Most common area
Ads and assets38Cases filed
Search terms27Cases filed
Keywords26Cases filed
Growth opportunities24Cases filed
Recent changes18Cases filed
Budgets and reach16Cases filed
Google’s recommendations14Cases filed
Structure and settings8Cases filed
Landing pages and speed6Cases filed
231 current cases on 10 advertisers’ boards, November 2026.

Ten audits, case by case

Open any account to see every case Claude filed, in the words the board used.

Home physiotherapy providerUnited Kingdom · GBP

21 of 22 conversions in 30 days were landing in secondary actions, so bidding was steering on a single conversion.

conversions bidding could not see
21 of 22
on one broad keyword, 0 conversions
£52.11
keywords blocked by their own negatives
7
What Claude found (6 cases)
  1. Conversion trackingThe actions that record are not the ones steering bidding

    The 8 primary actions recorded 1 conversion between them. The bookings that really came in were all counted as secondary.

  2. Recent changesEvery conversion action stopped at the same moment

    22 events in the 11 days to 21 October, then 0 in the next 11 days on 36 clicks. Claude flagged it as a tracking break, not a slow week.

  3. KeywordsA broad keyword spent more than a conversion costs and got nothing back

    £52.11 and 0 conversions on broad match, while the phrase copy of the same keyword won 2 conversions for £18.72.

  4. KeywordsSeven exact keywords were also added word for word as negatives

    Each one was bid on and blocked at campaign and ad group level, so it could never show.

  5. Ads and assetsThe ads promised same-day visits, a callout still said same-week

    Two promises side by side in the same ad, one of them out of date.

  6. Growth opportunitiesMondays converted for a fraction of the cost

    Monday brought 4 of the 8 conversions on under a tenth of spend: £7.48 per conversion against £38.24 across the week.

What happened next: The owner applied seven of the fixes from the board right after the audit, including pausing the broad keyword.

Multi-specialty medical clinicUnited Arab Emirates · AED

Specialist ads were sending AED 5,141 a month to a general health-checkup page instead of the specialist the ad promised.

in 30 days to the wrong landing page
AED 5,141
per extra conversion if Google’s raises were applied
AED 165
on Connected TV, 0 conversions
AED 404
What Claude found (5 cases)
  1. Landing pagesTwo specialist ads landed on a full-body checkup page

    AED 3,455 and AED 1,686 in 30 days went to a page selling a checkup, not the specialist service in the ad.

  2. Google’s recommendationsGoogle wanted to raise seven budgets

    Applying all seven would buy extra conversions at about AED 165 each. The account was running at AED 105.

  3. Demand GenA spending campaign reported no asset groups

    The one Demand Gen campaign showed zero asset groups while it had spent AED 5,655.74.

  4. BiddingTwo campaigns bid automatically below the learning floor

    Maximize Conversions was learning from 25.5 and 27.25 conversions a month, under the roughly 30 the strategy needs.

  5. AudiencesThe only Customer Match list could never serve on Search

    One list was uploaded, but its Search size was zero.

Local business with a store and lead formsIndia · INR

All 732 conversions in 30 days were map directions or call-button taps. The lead form and WhatsApp had recorded none.

conversions, all directions or call taps
732
actions set as primary, a test one included
12 of 12
conversions behind the campaign the spend limit stopped
596
What Claude found (5 cases)
  1. Google’s recommendationsThe main campaign was bidding toward taps, not leads

    Every conversion was a directions request or a call tap. Lead form, WhatsApp, and contact page recorded nothing.

  2. Conversion trackingSoft engagement actions and a test action were all primary

    All 12 actions fed bidding, including one named as a test and an engagement action that recorded 2,301 in a month.

  3. Budgets and reachThe account spending limit was used up

    Spend had reached the account limit to the rupee, so the campaign behind all 596 conversions was suspended.

  4. Performance MaxEvery conversion was worth exactly 1

    A placeholder value, so any return-based target would be meaningless.

  5. Growth opportunitiesSunday to Wednesday converted cheaper, yet the late week took most spend

    INR 36.01 per conversion early in the week against INR 47.47 Thursday to Saturday, which took 61% of spend.

Heating and air-conditioning serviceTurkey · TRY

One keyword with a Quality Score of 3 took 762.56 of the account’s 984.97 TRY and brought no conversions.

of spend on one keyword, 0 conversions
77%
for the landing page to show on a phone
66.6 s
TRY per conversion: one search vs the account
68 vs 787
What Claude found (5 cases)
  1. KeywordsOne poorly scored keyword took most of the spend

    Quality Score 3, 762.56 TRY of 984.97, no conversions.

  2. Search termsAds were paying for how-to questions

    Six “how” and “how often” searches took 223.47 TRY of 563.84 TRY in 30 days, with no conversions.

  3. Landing pages and speedThe landing page took 66.6 seconds to show on a phone

    Google’s bar for a good page is 2.5 seconds. Claude added a message to send to the developer.

  4. Recent changesA negative keyword blocked the company’s own name

    A broad negative excluded every search containing a word from the business name.

  5. Search termsA search that converted was not a keyword

    It brought one of the account’s two conversions for 68.39 TRY, when a conversion cost 786.80 TRY across the account.

Live-in home care agencyUnited Kingdom · GBP

Google’s auto-apply removed two keywords set up days earlier, and the main campaign nearly stopped showing.

days served after the keyword swap
1 of 5
night-care keywords blocked by their own negatives
7
live spend, 0 conversions; the paused campaign had 2
£204.69
What Claude found (5 cases)
  1. Recent changesAuto-apply removed two keywords

    Recommendations auto-apply removed two phrase keywords on 31 October that had been set up on 28 October.

  2. Recent changesThe main campaign nearly stopped after its broadest keyword was swapped

    Its last daytime impressions came the hour the keyword was removed. After that it served on 1 day out of 5.

  3. KeywordsA whole set of keywords was blocked by its own ad group

    7 enabled night-care keywords each matched a negative in the same ad group.

  4. Search termsA misspelled negative blocked nothing

    A list held “acility” instead of “facility”, so the singular search still got through.

  5. Budgets and reachThe only campaign that converted was switched off

    The paused campaign logged 2 conversions on £22.56. The live one logged 0 on £204.69.

Personal trainerBrazil · BRL

A keyword written as exact was running broad and matching gyms, studios, and people’s names.

different searches through one “exact” keyword
80
per conversion on the campaign losing 37% to budget
18.80 BRL
spelling mistakes in live ad copy
3
What Claude found (5 cases)
  1. Search termsOne keyword written as exact was running broad

    80 different searches came through it with no conversions, while the plain service keyword brought both conversions.

  2. Search termsExact-match negatives only blocked the bare word

    A price search slipped past the price negative, and gym searches past the gym one.

  3. TrackingThe final URL suffix held a full web address

    It should hold key=value tracking parameters; it repeated the site address on every click.

  4. Ads and assetsThe ad copy had spelling mistakes

    Three typos across a headline and two descriptions, one in the area name the ad is built around.

  5. Growth opportunitiesThe campaign converted and was still limited by budget

    37% of eligible impressions lost to budget, at 18.80 BRL per conversion.

Equipment repair businessSaudi Arabia · SAR

No conversion action had ever recorded anything, and the landing page HTML had no Google tag.

conversions recorded by 3 primary actions
0
spent in the first week with nothing measured
SAR 77
ad, so nothing to test it against
1
What Claude found (4 cases)
  1. Conversion trackingThe site may not have a tag at all

    All three primary actions recorded 0 in 30 days, and the page’s HTML showed no Google tag.

  2. Recent changesThe only campaign was paused and stayed off

    Nothing served for four days after a pause made in Google Ads.

  3. Ads and assetsOne ad, rated Average, carried every click

    With no second ad there was no way to tell whether the copy was the problem.

  4. Ads and assetsHalf the callouts Google recommends

    Three callouts, each repeating a sitelink, and no structured snippet.

Local emergency serviceSaudi Arabia · SAR

Phone calls were the goal, but the call asset had call reporting off and the tag had not recorded once in 12 months.

without a single recorded conversion
12 months
phone impressions in the month
0
misspelled headlines, one in the city name
2
What Claude found (3 cases)
  1. Conversion trackingNothing had ever been recorded

    The contact tag recorded nothing in 12 months while the campaign took clicks.

  2. Conversion trackingCalls were a goal but could not count

    The call goal was left out of the Conversions column, and call reporting was off on the newer call asset.

  3. Ads and assetsTwo headlines were misspelled

    One of the errors was in the name of the city the ads target.

Small online advertiserUnited Kingdom · GBP

One day recorded 70 conversions on £2.06 of spend, about 30% of the month, and made the account look cheaper than it was.

conversions on £2.06 in one day
70
per conversion: Tuesday vs the week
£0.20 vs £0.44
What Claude found (3 cases)
  1. Budgets and reachOne day distorted the month

    A single day carried about 30% of the 30-day conversions, so cost per conversion read lower than reality.

  2. Budgets and reachThe cheapest campaign was capped by its budget

    It spent its full daily ceiling at £0.29 per conversion against an account bar of £0.41.

  3. Growth opportunitiesEarly-week days converted at half the cost

    A Tuesday conversion cost £0.20 against £0.44 across the account.

Services companySaudi Arabia · SAR

All four enabled campaigns had 0 impressions in 30 days, and two Search campaigns were set to show on Display once they did.

impressions across 4 enabled campaigns
0
Search campaigns with Display switched on
2
What Claude found (4 cases)
  1. Growth opportunitiesNothing was serving

    SAR 46 a day of budget, 0 impressions and 0 spend in 30 days.

  2. StructureSearch campaigns would leak into Display

    Display was on in both Search campaigns. No Display spend yet only because nothing had served.

  3. TargetingAds would reach people only interested in the area

    Location option set to presence or interest, so searchers outside the area could trigger ads.

  4. Landing pagesThe landing page heading just said “Blog”

    Caught before any spend reached it, the cheapest time to fix it.

From real Boxoo audits, read in November 2026. We removed names, brands, cities, and identifying search terms; the numbers are exactly what Claude reported, in each account’s own currency.

What these audits teach

Check tracking before bids

In 7 of 10 accounts, the number bidding optimised toward was not the business result: secondary actions holding the real leads, map taps counted as conversions, or nothing recorded at all.

Negatives break keywords quietly

Four accounts had keywords blocked by their own negatives: 30 in one ad group, and in another a negative that blocked the business’s own name. Google Ads shows no error; the keyword just never serves.

Google’s advice needs your numbers

One clinic’s recommended budget raises would have bought conversions at AED 165 when the account paid AED 105. Auto-apply removed two fresh keywords in another account.

How each audit was done

The same process ran on every account above. Claude reads, checks its own numbers, and proposes only what the account can prove.

How Claude audits an account
  1. Reads the live accountThrough the Google Ads API: campaigns, search terms, keywords, ads, conversion actions, change history.
  2. Specialist reviews runEach review is a separate Claude run with one job, like search terms or tracking.
  3. Files cases with evidenceOnly what the numbers prove, in your currency, ranked by what costs most.
  4. Checks the fixA second pass confirms the numbers and the exact ids before it proposes a change.
  5. You decideApply in one click with an undo, or take the steps for anything outside Google Ads.

Your account could be the next case study. Connect Google Ads and read your own cases. Nothing changes until you press Apply.

Start my free audit
Free Chrome extension

Talk to Claude right inside Google Ads.

The Boxoo extension for Chrome is free. It puts the audit where you already work: a Boxoo score in the top bar, chips beside the ad groups, keywords, and ads that have a case, and a chat with Claude that already knows your account. Ask “why did my cost per conversion go up?” on the page you are looking at.

Chat with Claude in the Boxoo side panel, open on a Google Ads campaign page

Questions

Are these real Google Ads audits?

Yes. Every case study on this page comes from a real advertiser who connected Google Ads to Boxoo and got the free audit. We removed names, brands, cities, and search terms that could identify them. The numbers are exactly what Claude reported, in each account’s own currency.

What is the most common problem a Google Ads audit finds?

Conversion tracking. Seven of the ten accounts had at least one tracking case: real conversions counted as secondary, tags that stopped recording or never recorded, soft actions like map taps steering bidding, or every conversion carrying a placeholder value of 1. When tracking is wrong, smart bidding optimises toward the wrong thing, so it is checked first.

How much money does a Google Ads audit usually find?

It depends on the account, and we do not promise a figure. In these audits, single findings ranged from a few pounds on a misspelled negative to AED 5,141 a month sent to the wrong landing page. In a small account in Turkey, one keyword took 77% of spend with no conversions. Each case shows the exact amount from your own account.

Who does the checking?

Claude. Each area of the audit is a separate Claude review with one job: search terms, keywords, conversion tracking, budgets, ads, landing pages, targeting, recent changes, and Google’s recommendations. A second Claude pass checks the numbers and the exact campaign, ad group, or keyword ids before it proposes a fix.

Does the audit change anything in my account?

No. The audit only reads the account. A fix is applied only when you press Apply on a case, and Boxoo records the old values first, so most changes can be undone.

Can a small account get a useful audit?

Yes. Several accounts here spent under 1,000 a month in their own currency. Small accounts often have the clearest problems: one keyword taking most of the budget, tracking that never worked, or ads with spelling mistakes.

Is the audit free?

Yes. The Google Ads audit is free with no card. The Boxoo Chrome extension, which shows the audit and a chat with Claude inside Google Ads, is free to install.

How long does the audit take?

Connecting takes about a minute: sign in with Google, authorize Google Ads, pick the account. The reviews then run on Boxoo’s side and cases appear on your board as each one finishes, so you do not need to keep the tab open.

More on the free audit

Boxoo is not affiliated with Anthropic. Claude is a trademark of Anthropic, PBC.

See what Claude finds in your account.

The audit is free and read-only. Every case comes with the numbers and a fix you approve.