Google Ads in the United States: The 2026 Operator's Guide
Key takeaways
- The US is one of only two markets with a citable, methodology-documented Google Ads benchmark dataset: 13,474 US-based search advertising campaigns, 1 April 2025 to 31 March 2026, 23 industries, reported as medians described as averages.
- The same study appears in two places with different descriptions, and one says the sample includes Microsoft Ads campaigns alongside Google Ads. That is unresolved, and it matters for anything titled a Google Ads benchmark.
- Between-industry spread is roughly sixfold on cost per click. Within-industry spread is unpublished — the report gives medians and no dispersion measure, so it cannot tell you how far a normal advertiser sits from the middle.
- The all-industry US average CPC figures circulating in 2026 do not appear on the report page they are credited to, and the two common versions disagree with each other. Do not plan against either.
- US privacy runs on an opt-out model, so a browser-level opt-out signal has to be detected and acted on. That changes which tags fire, who enters a remarketing list, and what you may upload for Customer Match.
The short answer
The United States is the most competitive, most instrumented and most expensive search market in the world, and one of only two — with the United Kingdom — where a genuinely citable public benchmark dataset exists. That changes the problem. In most markets the honest advice is that the numbers you are reading were made up. In the US they are real, and the failure mode is subtler: real data, correctly measured, applied to an advertiser it does not describe.
So this post names the dataset precisely, then spends most of its length on why it still cannot plan your budget. For the opposite extreme, the Nepal benchmark investigation traces how figures get manufactured where no data exists at all — the same argument from the other end.
I have managed around USD 113,000 on Google Ads and USD 57,000 on Meta across seven or more countries, and the US is one of the markets I have actually worked in rather than only researched. The practitioner sections come from that; the benchmark sections come from reading the report, which is a different kind of claim and is labeled as one.
The dataset that actually exists
The reference most US paid search writing gestures at is the annual benchmark report published by WordStream, a USA TODAY Co brand, titled Google Ads Benchmarks 2026: Competitive Data and Insights for Every Industry. It was published on 19 May 2026 and has not been revised since. It deserves to be cited accurately.
| Field | Stated value | Why it matters |
|---|---|---|
| Sample | 13,474 US-based search advertising campaigns | Large enough that industry figures are not noise, and printed on the page |
| Period | 1 April 2025 to 31 March 2026 | A full year, so seasonality sits inside the figure |
| Type | Observed campaign performance, not forecasts | What advertisers were charged, not what a tool projected |
| Statistic | Medians, described in the report as averages | Outliers cannot drag the figure — but no spread is published |
| Segmentation | 23 industries, minimum 52 campaigns per subcategory | A floor on how thin any cell can be |
| Geography | United States only | Not a global benchmark, and never claimed to be |
Three representative figures from that edition, quoted with their industry labels because they mean nothing without them: Attorneys and Legal Services at an average CPC of $9.87 and a cost per lead of $131.63; Finance and Insurance at $3.39; Arts and Entertainment at $1.63 and a cost per lead of $26.84. Three, not a table, deliberately — a transcribed table invites you to find your row and stop thinking.
The ambiguity nobody flags
Here is the part left out of every citation I have seen, and it is disqualifying for the claim most people make with this data. The same study appears in two places under the same corporate family. The version titled as a Google Ads benchmark states the 13,474-campaign sample, the 23 industries, the April-to-March period, and that the figures are medians. The other version, published in June 2026, describes the sample only as thousands of customer campaigns, says "top 20+" industries, calls the figures averages, states no period — and describes the data as covering Google Ads and Microsoft Ads.
The industry numbers match across both, so it is plainly the same underlying data — but those descriptions cannot both be precise. If Microsoft Ads campaigns are in the sample, a study titled Google Ads Benchmarks measures something broader than its title says, and every figure quoted from it carries an unknown amount of Microsoft Ads. I could not resolve which framing is correct, and I will not pretend the question does not exist.
How to cite it honestly
Quote the more methodologically explicit version, state the sample and the period, and describe the figures as US search advertising benchmarks rather than as pure Google Ads benchmarks. That phrasing survives either resolution of the ambiguity. "The average Google Ads CPC in the US is X" does not.
A related warning: the two all-industry US average CPC figures that circulated in 2026 credited to this report appear nowhere on the report page, come from third-party summaries, and disagree with each other. Discard both.
Why a real national average is still nearly useless
This holds even if the data is flawless. Grant that the sample is clean and every figure exactly right — a national average CPC is still close to unusable for a specific advertiser, for three structural reasons.
The spread inside an industry dwarfs the spread between them
Across the three industries quoted above, cost per click runs from $1.63 to $9.87 — roughly sixfold. That is the between-industry gap everyone fixates on. Now look behind one of those figures: it is the median of at least 52 campaigns, run by advertisers with different customer values, geographies, match type discipline, landing page quality and bidding strategies, some bidding on emergency-intent queries and some on research queries filed under the same industry label. The distribution behind one median is wide, and wide in exactly the dimensions that determine what you pay.
The report publishes medians and no dispersion measure at all — no range, no quartiles, no standard deviation. That is not a criticism; almost no benchmark publishes one. But it means the data tells you where the middle of your industry sits and cannot tell you whether sitting at twice the median makes you an outlier or entirely typical. The second question is the one you actually had.
Geography inside the US varies more than some country gaps
The only cross-country cost index that exists is built from keyword planning forecasts rather than observed spend, and on it the United States is the baseline while comparable English-speaking markets sit within roughly 13 percent. That is the entire gap between the US and the UK on that measure. Hold it next to the US itself: the same keyword in the same vertical, bought in a dense coastal metro and in a thinly served rural county, competes against completely different advertiser sets with completely different customer values. If you believe those two auctions price within thirteen percent of each other, a national average is a reasonable planning input for you. Country is a weak predictor of price; location inside the country is a strong one.
Industry labels bundle intents that do not belong together
Every category merges things that behave differently. One healthcare row covers a high-value elective procedure and a routine appointment booking; one home services row covers an emergency callout and a planned renovation. The queries, urgency and auction depth are unrelated, and the resulting median describes no advertiser in particular.
What the benchmark is genuinely good for
- Order-of-magnitude sanity checks before entering a vertical. If your model needs a $2 click where the national median is close to $10, you have found a problem in the model before spending anything.
- Direction of travel. The report publishes year-on-year change, and a category-wide move is real market information in a way a level is not.
- Arguing for budget internally, because a finance team needs a reference point that did not come from the person asking for the money.
And three it will not support: setting your target cost per click, judging your agency, and deciding whether your account is healthy. Those need your own history and segments — the only populations you belong to.
The metric that lies in a US account
The most common way a US account degrades while looking fine is a falling cost per lead — the number everyone reports upward, the number the benchmark invites you to compare, and one that moves for two opposite reasons that look identical on a chart. It falls when the account gets better. It also falls when the definition of a lead quietly loosens: a chat widget interaction counted as a conversion, a phone click counted whether or not the call connected, a form counted on view rather than submission. Volume rises, cost per lead drops, and the sales team stops finding anyone worth calling.
The second number that catches it is the one nobody maintains: leads reaching a qualified stage, divided by leads generated, tracked weekly. If cost per lead falls and that ratio falls with it, the account is not improving — the counter is. The mechanics of counting conversions once and correctly are in the conversion tracking guide for GA4 and Tag Manager, and that is the prerequisite for every comparison in this post. Benchmarking a number you measured wrong only tells you precisely how wrong it is.
Local Services Ads — the format the US has and most markets do not
Local Services Ads operate in a small number of countries, and the US has by far the widest coverage — roughly 110 categories, and the only market with health care, legal, personal care, education, pet, food and financial verticals. Elsewhere the same format is a home services product with a short category list, so any playbook treating "run LSA for local service businesses" as universal advice is a US playbook that does not travel.
Some categories are restricted to particular states, and pre-badge ads — which serve while onboarding completes — are unavailable for several categories including health care and locksmiths. You are charged per lead, not per click, so treat it as a second channel with its own diagnostics.
Privacy, operationally — and only operationally
I configure ad accounts, tags, consent settings and audience lists. I am not a lawyer, this is not legal advice, and nothing below tells you what any law requires of your business — confirm your position with a qualified US attorney before changing anything. What follows is strictly what these regimes change about how an account is built.
The US model is opt-out, and that has a technical consequence
There is no single comprehensive federal privacy law in the United States. What exists is a state-level patchwork — roughly twenty states in effect in 2026, depending on whether one narrower regime is counted as comprehensive — and the common pattern is an opt-out of targeted advertising rather than an opt-in before tracking. That is the reverse of the UK arrangement, which is why a consent configuration built for one market is wrong in the other.
The operational consequence that matters most: a browser-level opt-out preference signal has to be detected and acted on where it applies. That is not a banner interaction — it arrives with the request, before anyone clicks anything, so your tag configuration has to read it and suppress advertising tags without waiting for a user action. Most sites that added a banner and called it done are not doing this.
What it does to remarketing and Customer Match
Two effects, often confused. Remarketing lists are built from tag activity, so every visitor whose advertising tags are suppressed never enters the list — lists grow more slowly than traffic and for smaller advertisers can sit below the size needed to serve. That is a design constraint, not a bug to engineer around: build fewer, broader segments rather than many narrow ones, and expect a shorter usable membership window.
Customer Match is different: the data comes from your records rather than a tag, so suppressing a tag does nothing for it. What governs an upload is the basis on which you collected and are using that contact data, and whether the individual has opted out of having their information shared for advertising. That is a records and legal question, not a Tag Manager question, and most accounts have never audited it.
Consent state changes which tags may fire
Your setup needs three behaviors, not one: full measurement where advertising is permitted, restricted measurement where it is not, and a defined default before anything is known. The failure I see most is a binary setup with no pre-decision state, firing everything for the first few seconds of every session. A delayed drop is not a consent mechanism, it is a slower one.
Lead-gen funnels that feed a call center
If a US lead form feeds a dialer or a call center, consent obligations attach to the outbound contact as well as to the tracking. What a person agreed to, when, and on which form has to be retained and retrievable per lead — so the consent language and a timestamp travel into the CRM as fields, not as a screenshot of the form somewhere. A common gap in otherwise well-run lead-gen accounts, and a question for your attorney rather than your marketer.
Billing and the cost lines outside the auction
US accounts bill in USD through Google LLC, and the US carries no jurisdiction-specific Google surcharge at all — several comparable markets do, one adding a standing percentage to every campaign. One recent change most published US material predates: per Google's own tax documentation, Washington State sales tax applies to digital advertising from 1 January 2026. Sales tax is state-level and varies, so verify against your own billing address in a live account rather than taking it from any article, including this one.
The prerequisite nobody wants to hear
None of this works if the numbers in your account are wrong, and in most accounts at least one is. Before benchmarking anything, establish that a conversion means one real outcome counted once: no duplicate conversion actions, no imported goals double-counting native tags, no form views counted as submissions, no phone clicks counted as calls. Each inflates volume and deflates cost per lead, making an account look better than the benchmark exactly as it gets worse. The Google Ads audit checklist runs the sequence in order, and it is the right first afternoon in any account you did not build yourself.
The decision test
When someone hands you a benchmark and asks whether your account is performing, work through these in order. The first clear answer is the answer.
- Is your conversion definition the same as the benchmark used? If you cannot state what counted as a lead in the study, stop — you are comparing two different events.
- Is your geography comparable? A national median against a single-metro account is not a comparison. Segment by location and see whether internal variance already explains the gap.
- Is your industry category the one you would actually have been filed under? Most advertisers sit at the edge of a category that bundles them with businesses selling something different at a different price.
- Does the difference survive segmenting by device, match type and brand versus non-brand? Brand traffic inside a blended average is the most common reason an account looks cheaper than its industry.
- Only if all four survive is the benchmark saying something about your account. It rarely gets past question one.
What this post is, and what to verify yourself
This reflects the position as of July 2026. Platform availability, tax treatment and privacy rules change, and several items above changed within the last twelve months. The benchmark report is annual and a new edition will supersede the figures here. Treat every specific as a claim with a date attached, and check the source before it enters a client plan.
On the legal material: I have described only what these regimes change about account configuration, because that is my competence and the boundary of it. Nothing here is legal advice, and anything with consequences goes to a qualified US attorney before it goes into production.
Where to go from here
Take the industry figure closest to your business, then spend an afternoon on the four things that decide whether it means anything: how you define a conversion, how your account segments by geography, how much of your volume is brand, and whether tracking counts each outcome once. Whatever comes out is a better planning input than any published median. The other market with real data works differently again — the UK guide covers a consent regime that changes what the account can measure in the first place.
That is the standard I hold the work to at Arcetis, the growth systems practice I run: no benchmark enters a client plan without its sample, its period and its definition of the thing counted attached to it.
Frequently asked questions
What is the average cost per click for Google Ads in the United States?
A real dataset exists — 13,474 US search advertising campaigns measured between April 2025 and March 2026 across 23 industries — but it reports by industry, not as one national figure, and the all-industry averages circulating in 2026 do not appear on the report page they are credited to. Two commonly quoted versions disagree. Use the industry figure nearest your business as context, and your own account as the number you plan against.
Are US Google Ads benchmarks reliable?
They are the most reliable in any market, which is a lower bar than it sounds. The sample is observed campaign spend rather than forecast, the period is stated, and there is a minimum campaign count per subcategory. What they do not publish is any measure of spread, so you see where the middle of an industry sits but not how wide the distribution is — and that missing number is the one you needed.
Do the US Google Ads benchmarks include Microsoft Ads campaigns?
Unclear, and nobody has resolved it publicly. The study appears in two places. The version titled as a Google Ads benchmark describes a sample of search advertising campaigns; the other, with matching industry figures, describes the sample as covering Google Ads and Microsoft Ads. Both cannot be precise descriptions of one dataset. Treat the figures as search benchmarks rather than pure Google Ads benchmarks until it is clarified.
Why is my cost per click higher than the US benchmark for my industry?
Usually because the benchmark is a median across an entire industry and a whole country, and you are one advertiser in one metro buying one keyword set. Industry categories bundle very different intents, geography inside the US varies enormously, and the figure carries no published spread. Segment your own account by geography, device and match type first — the variance inside your data usually explains most of the gap.
Do I need Local Services Ads in the US?
If you are a local service business in one of the roughly 110 US categories, it is worth testing: it is pay-per-lead rather than pay-per-click and sits above the standard results. It is not a substitute for Search — it covers a fixed category list, some categories are restricted to particular states or metros, and it requires verification before you run at full status. Treat it as a second channel with its own economics.
Does a browser opt-out signal affect my Google Ads remarketing lists?
Yes, operationally it does. Where an opt-out preference signal has to be honored, that visitor should not be added to advertising audiences, so remarketing lists grow more slowly than traffic and can fall below the size needed to serve. The answer is not a workaround — it is designing audiences that tolerate a smaller pool and leaning on first-party segments you have a clear basis to use. Confirm your position with a qualified US attorney.
Is Google Ads more expensive in the US than in other countries?
On the only cross-country index that exists, built from keyword planning forecasts rather than observed spend, the US is the baseline and comparable English-speaking markets sit within roughly 13 percent of it. That is a smaller gap than most people assume, and smaller than the variation between metros inside the US. Country is a weak predictor of what you pay; vertical, intent and location are strong ones.
Should US advertisers also run Microsoft Ads?
Test it if your traffic is desktop-weighted, and skip it if it is not. Tracked search referral data for June 2026 puts Bing at 8.73 percent in the US overall but only 2.30 percent on mobile. The US is the one comparable market where Bing keeps a non-trivial mobile presence — but it remains a desktop channel, and a mobile-heavy account will see little from it.
Book a free 10-minute call about your US account
If you run US search and want a second opinion on whether your numbers are actually comparable to the benchmark you keep measuring them against, bring the account to a free 10-minute call. I will tell you what I would check first, and if the honest answer is that the account is fine and you do not need help, I will say that instead.
Direct: +977 9846162626 · lamichhanesapun2@gmail.com
This post supports the frameworks documented in full on the Authority page.
Related reading