Every SEO gets asked this question eventually. The client sees a line item for branded search in the Google Ads invoice, works out that people typing the company name were going to find the site anyway, and asks why anyone is paying for that.
It is a fair question. I have given the opinion answer plenty of times myself, usually some version of “competitors will take the click if you stop”, which sounds right and proves nothing.
I have spent 20 years in ecommerce marketing and audited more than 1,300 online stores, working both sides of this argument, and I can count on one hand the accounts where somebody had actually measured it. The rest had a position, usually inherited from whoever set the account up, and a slide in the monthly deck defending it. So here is the test rather than the opinion. What the research says, how to set the measurement up, how to run the pause without wrecking it, and how to read what comes back.
This Is Not the Cannibalisation You Already Know
The word is overloaded, so let us separate the two problems before we go further.
The version SEOs deal with weekly is keyword cannibalization, where two of your own pages chase the same query and split the signals between them. That is an internal, organic-only problem, and the fix is editorial. Consolidate, redirect, pick a winner.
Brand search cannibalisation is a different animal. Nothing on your site is competing with anything else. Your paid result is competing with your free one, on a query where you already own the intent, using money. The two pages of the argument sit in different departments and often in different companies, which is exactly why nobody resolves it.
There is a reason it stays unresolved. Brand terms have low cost per click, high click-through rate and a wonderful return on ad spend, so brand campaigns make the whole account look good. Nobody rushes to switch off the campaign that props up the average.
What the Research Actually Shows
Two pieces of evidence are worth knowing, because you will need them in the room when someone claims this is settled.
Google published a meta-analysis of 390 of its own search ads pause studies, looking at what happens to organic clicks when the ads stop. The headline that gets quoted, that around 89 percent of ad clicks are incremental, is not the number that matters here. The number that matters is the breakdown by organic rank. When the advertiser already held the top organic result, only 50 percent of the ad clicks were incremental. At ranks two to four it was 82 percent, and below rank four, 96 percent.
Read that again with brand queries in mind. On your own brand name you are almost always the top organic result. Google’s own research puts you in the bucket where half the ad clicks are clicks you had already earned.
The second piece is more brutal. Economists at eBay ran a large-scale field experiment, switching off paid search on some engines while keeping others as a control, and published it in Econometrica. For brand keywords they found substitution was close to complete, with natural search catching almost all of the forgone traffic and attributed sales. Their broader finding, that non-experimental estimates of paid search returns are wildly optimistic because clicks and purchase intent are correlated, is the part I would tattoo on every reporting dashboard.
Now the caveats, because I want you to use this honestly. Both studies are a decade old, on click and sales data rather than profit, and the SERP has changed enormously since, mostly in ways that push organic results further down the page. Neither tells you what happens in your account. They tell you the prior you should start from, which is that on brand terms a large share of what you pay for is a toll on traffic you already own, and that the only way to know your share is to run the experiment yourself.
Instrument the Question Before You Touch Anything
Skipping this stage is why most brand-pause tests produce an argument rather than an answer. Give it a fortnight before you pause a thing.
Define the brand query set, precisely. Open Search Console, go to the Performance report, and build a regex filter that catches the brand name, the common misspellings, the domain typed as a query, and the spaced and unspaced variants. Something in the shape of (brandname|brand name|brandnam|brandname\.com). Save it. Every number in this test depends on both platforms counting the same queries, and eyeballing “queries containing our name” will not do it.
Split brand into two buckets. Pure navigational queries, where somebody types the brand and nothing else, behave completely differently to brand plus modifier queries like “brand reviews”, “brand discount code”, “brand vs competitor” or “brand size guide”. You may own position one for the first bucket and sit fifth behind an affiliate roundup for the second. Those two buckets can honestly deserve opposite decisions, and averaging them together is how teams end up with a result that feels wrong to everyone.
Then go looking for every campaign that touches brand traffic, because this is where audits go sideways. Brand queries leak into Performance Max, Demand Gen, dynamic search ads and broad match non-brand campaigns, so pausing “the brand campaign” often pauses a fraction of the brand spend. Brand exclusions on Performance Max and Demand Gen are the control that actually holds. Add them, then verify with the search terms report a few days later rather than trusting the setting.
Seasonality is the next trap. Pull 13 months of brand query data so you can see the annual shape, then at minimum a stable 4 weeks immediately before the test, and mark every event that moves brand search on a calendar. Email sends, a TV or podcast spot, an influencer post, a product launch and a sale all spike branded queries. Run the pause across a promotion and you have not measured anything, you have just watched a busy fortnight.
Last, write down the numbers you will compare, in advance. Paid brand clicks and cost. Organic brand clicks and impressions. Organic CTR on the pure-brand query set. Branded revenue and orders. Total brand clicks, paid and organic added together, which is the number the whole test turns on. Deciding the metrics afterwards is how a null result quietly becomes a win.
Two extra checks before you go. Run Auction insights on the brand campaign to see who else is bidding on your name, because a competitor sitting above you changes the whole calculation. And note your mobile SERP, since a brand query with shopping ads, sitelinks and an AI answer above the fold is a different page to the one you see on desktop.
Running a Brand-Pause Test That Holds Up
There are two designs worth using, and the one you pick depends on how much traffic you have.
The cleaner design is a geographic holdout. Split your market into two comparable sets of regions, keep brand ads running in one and switch them off in the other, and compare the change between groups over the same weeks. Because both groups experience the same promotions, seasonality and algorithm updates, you get a genuine control. It needs enough brand volume in each region to see past the noise, which rules it out for smaller advertisers.
The fallback is a time-based on-off design. Two weeks off, two weeks on, repeated at least twice. Alternating cycles matter more than length here, because a single before-and-after comparison cannot separate the pause from whatever else happened that month. If you only ever run one two-week pause, be honest in the report that you measured a period, not an effect.
Whichever design you use, freeze everything else. No budget shifted into non-brand campaigns, no new landing pages on brand destinations, no changes to sitelinks or promotions, no rebrand, no big email push, no PR launch. If the marketing calendar cannot give you four undisturbed weeks, take the four weeks you can get and note the contamination rather than pretending it away.
Set a stopping rule before you start. Mine is simple. If a competitor’s impression share on brand queries rises materially mid-test, or organic brand clicks fall while total clicks fall with them, the test ends early and the ads go back on. You are running this to find out. If you catch yourself hoping for a particular result, that is worth noticing.
If organic brand clicks do slide mid-test, rule out the ordinary causes before you blame the pause. The usual method to investigate ranking drops using Google Search Console applies here unchanged, and a site-wide ranking problem that happened to land in your test window will otherwise be recorded as an advertising finding for the next two years.
Reading the Result
Here is the arithmetic, which is less complicated than the debate around it.
Take the paid brand clicks you gave up. Take the organic brand clicks you gained over the same window against baseline. The recovery rate is the second divided by the first. Whatever is left over is your incremental click count, the traffic that genuinely disappeared when the ads went dark.
Then price it. Divide the spend you saved by the incremental clicks lost, and you have what those clicks were really costing you, which is always dramatically higher than your reported brand cost per click. A brand campaign at 40 cents a click with a 70 percent recovery rate is not a 40 cent click. It is a dollar thirty-three for the clicks that were actually additional, and you should compare that against your non-brand cost per click before deciding it is cheap.
Now do the same on revenue, because clicks are not the deliverable. Compare branded revenue and order count, paid and organic combined, against baseline. This is where a lot of tests fall apart, since attribution moves traffic between channels when you pause ads, and last non-direct models will shuffle credit around without any real behaviour changing. Look at total branded revenue rather than channel-level revenue, and treat direct traffic as part of the brand bucket, because a chunk of it is brand search that lost its referrer.
A worked example with made-up round numbers, purely to show the shape of the calculation. Baseline month, 10,000 paid brand clicks at 40 cents, so $4,000 spend, plus 30,000 organic brand clicks. Pause month, zero paid clicks and 36,000 organic brand clicks. You recovered 6,000 of the 10,000, a 60 percent recovery rate, and lost 4,000 clicks that were genuinely incremental. That $4,000 was buying 4,000 additional clicks at a dollar each, not 10,000 clicks at 40 cents. Whether a dollar is a good price depends entirely on what those clicks are worth, which is why the revenue comparison matters more than the click comparison.
One statistical warning. Brand click volumes wander week to week on their own, and small accounts wander a lot. If your swing is inside about 10 percent and you have only run one cycle, you have measured noise. Run more cycles or accept that the honest answer is “we cannot tell at this volume”, which is still more useful than an invented one.
The Cases Where Paying for Your Own Brand Is Correct
I am not arguing for switching brand campaigns off. I am arguing for knowing. Several situations genuinely justify the toll, and a good test surfaces them rather than burying them.
Competitors bidding on your name is the obvious one. If Auction insights shows a rival appearing on half your brand impressions, your paid result is not buying a click you already had, it is denying a click to someone else. That value is real and it never shows up in a recovery-rate calculation, so judge it separately.
Weak organic ownership of the brand plus modifier bucket is the second, and it is far more common than people expect. Affiliate roundups, review aggregators, marketplace listings and resellers routinely outrank the brand itself on “brand reviews” and “brand discount code”. If you are not position one there, Google’s own rank breakdown says most of those ad clicks are incremental, and the ads are doing exactly what ads are supposed to do.
Paying is the short-term answer in that situation. The durable one is taking the query back, which usually means building the page that answers it and giving it the internal links and structured data to compete, across a catalogue rather than one URL at a time. That is bulk work, better suited to tooling than to anyone with a spreadsheet, and it is the sort of project that quietly retires an ad spend line twelve months later.
Then there is control of the message. A sale, a shipping cut-off, a recall, a rebrand, a stock shortage. The ad is the only element of that SERP you can change this afternoon, and during a period when the message matters more than the click economics, that alone can justify the spend.
Finally, watch the pure navigational bucket in accounts running heavy paid social or offline media. Those channels drive brand searches, and cutting brand ads during a big awareness push can hand the last click to a comparison site at the worst moment.
Reporting It Without Starting a War
The internal politics here are worse than the mathematics. The person running the ads account has a monthly efficiency target, and your test threatens the campaign with the best numbers in the deck. Approach it as a shared experiment, not an audit finding, or you will get compliance instead of cooperation and the test will quietly not happen.
Two things make it land. Give the paid team the incremental cost per click figure rather than a verdict, because it is a media-buying metric they can act on and it lets them make the call themselves. And commit in advance to reporting the result either way, including the outcome where the ads turn out to be pulling their weight and you were wrong.
In the report itself, keep the experiment log visible over time rather than burying the result in one month’s commentary. If you already maintain SEO reports for clients with a fixed structure, add a standing experiments section, since this is the kind of finding that gets forgotten and re-litigated every time a new marketing manager arrives. The same applies to the SEO KPIs and dashboards you already track. Total brand clicks, paid and organic combined, deserves a permanent tile, because it is the number that catches this problem and almost nobody plots it.
One more habit worth building. Recheck the answer annually, or after any rebrand, SERP layout shift or new competitor. This is a measurement with a shelf life, not a permanent policy.
Five Ways These Tests Get Botched
Most failed brand-pause tests fail the same handful of ways.
- Pausing the brand campaign while Performance Max, Demand Gen or a broad match campaign keeps serving on brand queries. You measure a partial pause and conclude the ads did nothing.
- Running the test across a promotion, a launch or a big email send, then attributing the branded traffic spike to the experiment.
- Comparing paid clicks against organic sessions from two different tools with two different definitions of a brand query. Same definition, both platforms, or the numbers do not mean anything.
- Measuring channel revenue rather than total branded revenue, so attribution reshuffling looks like a real change in customer behaviour.
- Calling a single two-week window a result. Brand search moves on its own. One cycle tells you what happened, not what your ads do.
If you are already working through reasons your SEO is not working or the common SEO mistakes that quietly drain performance, add this one to the list. Not because paying for brand terms is always wrong. Because almost nobody knows which case they are in, and finding out costs a fortnight of discipline. Next time the question comes up in a meeting, it would be good to answer it with a figure out of your own account.
About the Author
Josh Uebergang is one of the most experienced SEOs in ecommerce. Over 20 years he has audited more than 1,300 online stores, recovered brands from traffic collapses most agencies walk away from, and built the SEO software thousands of merchants now run on their catalogues. He founded Digital Darts, the Shopify SEO agency the platform’s fastest-growing brands call when organic revenue has to move.