Stop paying for keywords you already rank for

Your paid manager could name their ten most expensive keywords from memory. Your SEO lead could name the ten they rank best for. At northfield.example nobody had put the two lists on the same page in two years. The same terms sat near the top of both.

That overlap is the expensive question neither report answers on its own. Which keywords are you paying for that organic would have caught anyway?

Answering it takes a join. Your Google Ads query set, with its clicks, cost and conversions per term, lined up against your organic rank for the same term. Conversions here mean the named goals your team defined in your analytics tool, a Purchase or a Demo request.

Which keywords are you paying for twice?

The ones where you already rank near the top and buy the click anyway. One check runs the join: paid performance per keyword on one side, organic clicks and average position from Search Console on the other, matched on the term. The flag goes up where both are true. That is where a paid click most likely replaced a free one.

A flag is a candidate and nothing more. Overlap shows that both channels touch the term. It cannot show you what happens when the ad disappears, and reading the flagged list as a cut list skips the step that makes the whole thing trustworthy.

Proof of substitution needs a controlled pause. That means a designed test with a holdout, read at the same significance bar as everything else you report, so the change has to beat ordinary week-to-week wobble before anyone calls it real. Suppress the ads in one slice. Leave a comparable slice running. Read the difference.

The same sheet carries what that test will need later. What you pay for the term month to month. The conversions attached to it, in your own named goals. And the organic position it already holds. A term with strong paid conversions is a different conversation from one where paid spends and nothing converts.

See also: what a defensible holdout looks like →

A brand outage ran the experiment

Our cleanest evidence on this arrived uninvited. A paid brand campaign went down for a stretch, unplanned, and the experiment nobody would have signed off on ran itself until somebody noticed.

Organic search picked up a large share of the clicks the paused ads had been collecting. Brand demand did not vanish when the ads did. It walked in through the unpaid door.

Then the arithmetic reframed the program. Once you count the organic pickup, the true cost of each incremental paid brand click lands at several times the sticker price, because most of the clicks being bought were arriving anyway. The question got sharper. Which brand clicks are actually incremental, and what are you paying for the ones only paid could win?

What the flagged list looks like

The player below shows the shape of the answer: the flagged terms, the spend attached to each one, and the caution that keeps the list from reading as a decision somebody already made.

Read the scope line before you read the numbers. Ads and Search Console finalize on different lags, so the join uses the window where both are complete and prints the as-of date beside the answer. That is how a lag stops being mistaken for a loss.

Why brand terms got harder to cut

This is harder than the classic version. We watched a paid team meet all three parts of it inside one quarter. Brand clicks declining overall. Competitors bidding on the brand name. And AI Overviews, the AI-written block Google puts on top of a classic results page, absorbing attention organic used to win.

Pausing brand spend in that weather is a bigger bet than it was when this check was invented. Their answer was the right shape. Keep defending auction presence in the short term, and commission the incrementality test rather than argue from theory.

One more modern complication deserves naming. Campaign types that hide query-level data make the join blurrier, so this check is sharpest where search terms are still visible, and that visibility is worth defending in your account structure.

Split branded from non-branded before you read the flagged list. Branded overlap is a question about defense economics and what a competitor's bid changes. Non-branded overlap is the cleaner arbitrage, where a strong organic rank and a converting page absorb demand at no marginal cost while your paid dollars move elsewhere.

What do you do with a flagged keyword?

It depends on where you rank. The policy is cautions rather than bans, because brand competition, seasonal pushes and auction dynamics all justify an override. What this check contributes is the starting position for each term, refreshed monthly, so your overrides are choices rather than defaults nobody revisited.

The tiered paid and organic operating policy article carries the full version. The short version below reads straight off the organic position you already hold.

• Strong organic position: trim paid back to a defensive floor and watch what organic does.
• Mid-page position: let paid accelerate while the organic work climbs.
• No organic presence at all: paid carries the term outright.

The workflow that does this: Paid query losses →

One place, one monthly conversation

Two source pages describe the two halves of the join. The Google Ads one covers paid, the Search Performance one covers organic. This check exists only because both live in the same place, which is the quiet argument for one shared data lake over two reports.

So run it monthly. Hand the flagged list to whoever owns the test calendar, and let brand terms graduate to a proper holdout once the spend justifies the window.

The effective cost article picks up the arithmetic from here: what a brand click truly costs once you count the organic pickup. The outage taught its lesson for free. A test is how you learn the same thing on purpose.

See also: what a brand click costs after organic pickup →