There is a report every SEO eventually builds and every client eventually loves: the overlap report. Keywords where you rank organically and buy ads. Multiply the ad spend on those terms, present the total, call it waste, recommend pausing. It is a satisfying deliverable. It makes you look sharp, and the saving is immediate and attributable.
It is also, in a meaningful share of cases, advice that loses money — and the mechanism by which it loses money is well understood and almost never mentioned in the posts recommending it.
This is not an argument that overlap does not matter. It is an argument that overlap is a hypothesis generator, not a decision, and that treating it as a decision is the actual expensive mistake.
The word the overlap report is missing
The overlap report answers: *where do paid and organic both appear?*
The question you needed answered: *if I turn this ad off, do I keep the conversion?*
Those are different questions, and the first is a poor proxy for the second. The concept that connects them is incrementality — the share of conversions that would not have happened without the ad. Overlap measures co-presence. Incrementality measures causation. Co-presence tells you almost nothing about causation.
The overlap report implicitly assumes that a click on your ad, when you also rank organically, would have arrived organically instead. Sometimes true. Sometimes very false. And the report contains no information whatsoever about which case you are in — that information simply is not in the data it uses.
Why "you rank #1, so pause the ad" is a 2019 argument
Here is the part that makes the classic advice genuinely obsolete rather than merely imprecise.
Organic position 1 used to mean *the first thing on the page*. In 2026, organic position 1 can mean: below four ads, below an AI Overview that answers the question outright, below a People Also Ask block, below a product carousel. Your "number one ranking" can sit most of a scroll down the page.
The ranking is identical. The visibility is not remotely identical. And every overlap report ever built reads the ranking.
This is why the overlap report and reality diverge most severely exactly where the money is: high-commercial-intent queries, which are precisely the queries with the most ads, the most features, and therefore the deepest organic results. The report is least trustworthy where it matters most.
The quadrant that replaces the overlap list
Stop sorting by "do we rank?". Sort by two axes that together describe whether your organic result is actually reachable:
- Axis 1 — Organic pixel position. Not rank. How far down the page is your result, on the SERP layout that is actually being served, on the device that actually matters for this query.
- Axis 2 — SERP capture. How much of the visible above-fold real estate belongs to someone other than you: ads, AI Overview, competitor features, carousels.
| Low SERP capture (you own the fold) | High SERP capture (fold belongs to others) | |
|---|---|---|
| Organic above the fold | Test pausing. This is the only quadrant where the classic advice is likely right. Your result is the first thing seen. The ad is plausibly buying a click you would get free. Test it — do not just cut it. | Hold and investigate. You are visible but crowded. Pausing may be survivable, but the ad is defending a contested fold. Test with a small holdout, not a full pause. |
| Organic below the fold | Keep the ad. Rare quadrant — usually means few ads but a heavy feature block. Your organic result exists but is not being seen. The ad is your visibility. | Keep the ad, and this is the trap. You "rank #1", the overlap report screams waste, and pausing donates the click to a competitor. This quadrant is where the standard advice does the most damage. |
The bottom-right cell is the entire point of this article. It looks identical to the top-left cell in an overlap report — both say "we rank #1 and we are paying for ads" — and the correct action is exactly opposite. Any report that cannot distinguish them is not helping you decide anything.
Brand terms: the argument that never ends
Nowhere is this fought harder than branded search. Someone searches your company name. You rank #1, obviously. You are also paying for the ad. This looks like the purest waste in the entire account.
The honest answer is: it depends on the SERP, and you can measure which case you are in cheaply.
Pausing brand ads is likely fine when: no competitor is bidding on your name, the SERP is uncluttered, your organic result plus sitelinks own the fold, and the query is unambiguously navigational — the user is trying to reach you, not evaluate you.
Pausing brand ads is likely expensive when: competitors are actively bidding on your name (they will simply take the slot you vacated), your brand is ambiguous with another entity, the SERP carries comparison or review features, or the query is `brand + review` / `brand + alternative` / `brand + pricing` — where the user is still deciding and the fold is a battlefield.
The only method that actually answers the question
Everything above narrows *what to test*. None of it tells you what to cut, because you cannot infer causation from an overlap table no matter how good the table is. The only reliable instrument is a holdout test, and it is more accessible than most teams assume.
The geo-split holdout
- Choose two sets of geographic regions with similar traffic volume, similar conversion rate, and similar seasonality. Match on trailing behaviour, not on population.
- Pause the ad group in set A. Leave set B running untouched.
- Run it long enough to cover your full conversion lag plus a stable baseline. If your sales cycle is three weeks, a one-week test measures nothing.
- Measure total conversions — paid plus organic, combined — in both sets. Not organic lift. Not cost per click. Total.
- The difference between A and B, normalised against their pre-test baselines, is your incrementality.
Everything hinges on step 4. If you measure only organic lift, you will always see organic rise when ads pause, and you will always conclude the ads were cannibalising. That conclusion is unfalsifiable by construction — organic rising is guaranteed regardless of whether total demand fell. The number that matters is whether the total held.
Wrong question: did organic conversions go up when we paused?
(yes — almost always — and it proves nothing)
Right question: did TOTAL conversions hold when we paused?
Total held → the ad was buying clicks you had anyway. Cut it.
Total fell → the ad was incremental. Pausing it cost you money.
Total fell less
than the spend → the ad was inefficient but not worthless.
Now it is a bidding question, not a pause question.That third outcome is the most common and the most ignored. The answer to "this ad is inefficient" is usually a bid adjustment, a match-type change, or a negative keyword — not a pause. Pause is the bluntest instrument available and it is the one the overlap report always reaches for.
When you cannot run a clean geo-split
Small accounts, single-market businesses and long sales cycles make geo-splits impractical. Two weaker but usable alternatives: time-based on/off cycling over enough alternating periods to average out weekday and seasonal effects (noisier, needs patience, and is invalidated by any concurrent change), or Google Ads' own conversion lift studies if your volume qualifies. Both are worse than a geo-split. Both are dramatically better than an overlap report.
Where the real waste usually is
Having spent this long arguing that overlap is not waste, it is worth being concrete about what is. In most accounts the genuinely recoverable money is not in the overlap at all:
- Broad match bleeding into informational intent. You are paying commercial CPCs for people researching a definition. They will never convert, on any channel. This is real, unambiguous waste and it is invisible in an overlap report.
- Queries where an AI Overview answers completely. Paying for a click on a question the SERP already resolved above your ad. Low intent, high CPC.
- Terms you rank for and cannot convert on. If organic traffic to that page converts at near zero, buying more of the same traffic will not fix it. The page is the problem, not the channel.
- Competitor terms with no differentiation story. Buying "competitor alternative" clicks that land on a generic homepage. The spend is real; the message is not there to meet it.
- Duplicate coverage inside the ad account. Two campaigns bidding against each other on the same query. This is the one true "cannibalisation" in the story and it is entirely internal to paid.
Notice what these have in common: each is a mismatch between what the searcher wanted and what you paid to show them. That is a real, diagnosable problem. "We also rank organically" is not a problem — it is a coincidence that might be a problem, and might be the opposite.
What to do on Monday
- Build the overlap report anyway. It is a fine starting list. Just relabel it — it is a *test queue*, not a cut list.
- Enrich it with SERP data: pixel position, feature composition, above-fold ownership. This is the step everyone skips and it is the step that makes the report mean something.
- Place each term in the quadrant. Only the top-left is a pause candidate. The bottom-right is a keep, no matter how much the spend column offends you.
- Segment brand into navigational and evaluative and stop arguing about the composite.
- Test the top-left with a geo holdout, measuring total conversions across both channels.
- Act on the test, not on the report. And when the test says the ad was incremental, say so out loud — that finding is worth more than the saving you were hoping to present.
The reason this is worth the extra work is not that it saves more money. Sometimes it saves less. It is that it produces decisions that survive contact with the following quarter. An overlap-driven pause looks brilliant for one month and then quietly shows up as a conversion shortfall that nobody connects back to the cut — because the saving was measured in one channel and the loss landed in another.
That disconnect is the whole problem. Measure the total, and it goes away.
SEMOptimiser's campaign tracker joins your ad spend to organic position *and* SERP feature composition on the same query, so the quadrant is built for you rather than assembled by hand — but the holdout test is still the only thing that answers the question, and no tool can do that part for you.