A $2,000 day on a $1,000 daily budget can sit fully inside Google's rules. So can a campaign that pays for clicks with no chance of becoming customers. Neither fact tells you whether your account is wasting money, and this audit closes that gap.
After Alphabet's Q3 results, advertisers raised three concerns: irrelevant clicks from broader keyword matching, campaigns spending up to twice their daily budget, and AI placements that can't be reported separately. According to Search Engine Journal's coverage, the earnings numbers don't prove Google changed its auction or budget rules to pay for AI. You can't settle that debate from your desk. You can find out whether your own spend meets your own targets.
Set the standard before you open the account
Waste only exists relative to a target. Write yours down first, or every number you pull will look fine or alarming for no good reason.
You need three figures:
- The most you can pay per conversion, or the ROAS you need.
- The conversion that counts as real value, such as a sale or qualified lead rather than a button click.
- A date window long enough to include your normal conversion lag. Thirty to ninety days works for most Search accounts.
Use the same window for every step below.
Step 1: Audit the search terms report
Irrelevant clicks show up here. Export the search terms report for the full window, with cost, clicks, conversions, and conversion value.
Classify what you paid for
Sort by cost, then label each high-spend term as relevant, borderline, or irrelevant. Judge by intent: would a person typing this query ever buy what you sell? Start with the top terms by spend, since a long tail of tiny terms rarely changes the picture.
Calculate your irrelevant spend share
Divide the cost of irrelevant terms by total cost. This one number tells you more than any anecdote about "bad traffic."
As a working heuristic (not a Google rule), also flag any term that has spent several times your target cost per conversion with zero conversions. Treat those as candidates for negative keywords, not automatic cuts, because small samples mislead.
Step 2: Break results down by match type
The broad-match complaint is testable. Segment performance by match type, using the match type column in the search terms report where your account shows it and the keywords report for keyword-level data. We cover related ground in how to validate an affiliate marketing niche in 5 steps explained.
Compare three things for each match type: cost per conversion, conversion rate, and irrelevant spend share. If broad match meets your target, the complaint doesn't apply to you. If it misses while phrase and exact match hit the target, you have a specific, fixable leak.
Step 3: Check budget pacing against Google's actual rules
Search Engine Journal reports that Google defines daily budgets as averages, and that most campaigns can spend up to twice that amount in a single day, subject to monthly limits (Search Engine Journal). A 2x day alone isn't a violation or a billing error.
The real question is whether the extra spend was productive. Pull daily cost by campaign and mark every day above your daily budget. Compare cost per conversion on those days with all other days. Then confirm the monthly total stayed within your daily budget multiplied by the days in the period.
Step 4: Review bidding and conversion quality
A bidding strategy optimizes toward whatever you count as a conversion. Count weak actions, and the system will efficiently buy weak outcomes.
Check which conversion actions feed bidding and whether any low-value actions inflate the totals. Compare actual cost per conversion with the target you set on the strategy. Then compare platform conversions with your CRM or sales records for the same period. If leads look cheap in Google Ads but few become sales, your waste sits in lead quality, and no search term filter will fix it.
Step 5: Infer AI impact without pretending to measure it
Google currently combines AI Mode performance with standard campaign and search terms reporting. You can assess overall results, but you cannot cleanly compare AI Mode's acquisition cost, conversion rate, or lead quality against other Search traffic. Google has cited similar click-through rates for sponsored and organic product listings in AI Mode, but that doesn't show how those visitors convert (Search Engine Journal).
So the audit can only infer. Look for a date when cost per click rose, conversion rate fell, or search terms became longer and more conversational. Then check whether the shift concentrates in broad match or in particular campaigns. See a closer look at affiliate site audit: measure your google traffic risk for additional background.
Treat that pattern as a lead worth testing, not evidence that AI placements caused it. Seasonality, competitor changes, or your own edits fit the same data.
Worked example (hypothetical, for illustration only)
These numbers are invented to show the method. They do not describe any real account or benchmark.
A campaign has a $1,000 daily budget and a target cost per lead of $80. Over 30 days, it spends $29,450 and produces 370 leads. The review runs like this:
- Search terms: Terms labeled irrelevant account for $4,120 of spend and produce no leads. Irrelevant spend share is $4,120 divided by $29,450, about 14%.
- Match types: Broad match spent $15,000 for 150 leads ($100 per lead). Phrase spent $9,000 for 140 leads (about $64). Exact spent $5,450 for 80 leads (about $68). Of the $4,120 in irrelevant spend, $3,300 sits in broad match.
- Budget pacing: Four days hit $2,000. Those days spent $8,000 for 70 leads (about $114 per lead). The other 26 days spent $21,450 for 300 leads (about $71.50 per lead). The monthly total stayed below $30,000.
- Reading the result: The 2x days were permitted, but they missed the $80 target. Broad match is the likeliest suspect for both the irrelevant clicks and the expensive days. To confirm, check whether the 2x days concentrate in broad match.
In this scenario, the sensible move is to tighten broad match with negatives and test it against a narrower setup. Concluding that Google changed something would go beyond the data. Phrase and exact match meet the target and need no changes.
Decide: tighten, test, or leave alone
Tighten when a segment misses your target across a meaningful sample and the search terms confirm the traffic is off-intent. Add negatives, restructure, or reduce budget on that segment.
Also read: see how to label free articles on a paywalled site: test plan
Test when results are borderline, the sample is small, or your sales cycle is long enough that conversions may still be arriving. Run a controlled change instead of a blanket cut.
Leave alone when a segment meets your cost and quality targets, even if its CPCs look high. A high click price alone doesn't show that something is broken. Search Engine Journal notes that Google's published auction rules include Ad Rank thresholds, which can produce substantial CPCs without another eligible ad directly below yours.
Verify your findings before you trust them
An audit produces hypotheses. Confirm each one with a before/after comparison.
Change one thing at a time and annotate the date. Compare equal-length windows covering the same days of the week, and let conversion lag pass before you judge. Where you can, use a campaign experiment or a split by campaign so a control group runs alongside the change.
Track the same metrics you audited: irrelevant spend share, cost per conversion by match type, cost per qualified lead from your CRM, and cost per conversion on days above budget. If the change improves them against your targets, keep it. If not, revert.
What this audit cannot show
Three limits matter:
- Google combines AI Mode data with standard reporting, so no step above isolates AI traffic.
- Nothing here proves Google changed auction or budget rules. Search Engine Journal's report says the earnings figures don't establish that.
- No audit can promise lower costs or better performance, because results depend on your market, your offer, and your conversion tracking.
What to expect next
Your first pass will produce a short list of suspect segments, not a verdict. Re-run the same audit monthly with the same targets and window so changes in your own data become visible over time. If Google later separates AI Mode reporting, the match type and pacing baselines you built will let you compare it against your existing numbers quickly.
