What Google Ads costs, and what decides whether it pays
Google Ads is an auction. You bid on a search phrase, you pay when someone clicks, and whether that turns a profit depends almost entirely on which phrase you picked. The average cost per click across all industries was $5.26 in 2025, and that average hides the range that actually decides your budget.
In Google’s keyword planning data for the United States on August 3, 2026, a click on “plumber near me” costs $36.62. “hvac repair” runs $32.44, “roof repair” $37.22, “insulation contractor” $18.59, “mobile welding” $5.62. At the top of the market, “personal injury lawyer” costs $163.25 per click, with top-of-page bids reaching $313.64. A welding shop and a personal injury firm cannot run the same playbook, and no general claim about whether Google Ads “works” survives that spread.
So before you fund anything, run one calculation. Take your cost per click, divide by your realistic landing page conversion rate, then divide by your close rate. That is your cost per acquired customer. If it lands above what a customer is worth to you, no amount of bid tuning fixes it. The offer or the landing page is the problem.
If you are still deciding whether paid search deserves the spend at all, which marketing channels truly drive your sales is the better place to start, and my Google Ads strategies breakdown goes further into tactics.
What are Google Ads?
You are buying timing. Somebody types “hvac repair” into a phone at 11pm in July because the air conditioning just quit, and paid search is the one channel where you can stand in front of that person at that second. That is what the click is for. It is also why the click costs what it costs.
The mechanism itself is simple. Google Ads, called AdWords until 2018, runs an auction on search phrases and places the winners on search results and across a network of partner sites, apps, YouTube and Gmail. You pay when someone clicks, which is where pay-per-click comes from. No click, no charge.
Targeting goes well past the keyword: location down to a radius, language, device, hour of the day, day of the week, and audience signals layered on top. Formats cover text ads on search, image and video ads across the Display Network and YouTube, shopping listings, and app installs.
Why do so many businesses stay in it?
Start with the number that is beyond dispute. Google’s advertising revenue came to $294.7 billion in 2025, according to Alphabet’s fourth quarter and fiscal year 2025 earnings report (PDF), with Google Search and other up 13.4% for the full year. That is not evidence about your account. It is a measure of how many advertisers keep renewing.
On the advertiser side, WebFX’s roundup of Google Ads statistics cites Google’s own estimate of an 800% return on investment. Treat any platform-wide average as a number with an enormous spread around it, not a forecast. Reach is genuinely large: Marketing LTB’s 2025 statistics roundup has Google Ads touching over 90% of internet users across more than 2 million sites and apps.
The cheapest win in the platform is remarketing, which shows ads to people who already visited your site and left without doing anything. It is cheap for the obvious reason that the audience already knows who you are. I get a remarketing audience collecting before I turn on a single search campaign, so there is a list waiting by the time traffic arrives. My detailed guide to retargeting ads walks through the setup.
How do you open an account?
Sign in at ads.google.com with a Google account you already own and Google walks you through goals, business details and billing. Fifteen minutes, and most of it is billing.
Then wait for advertiser verification. Google starts it with an in-account notification or an email, and Google decides which tasks you get: a government-issued photo ID, your organization’s registration documents or D-U-N-S number, a six-digit code sent to the phone number on the account, and in the United States a Social Security number. You do not pick a method, and there is no postcard. Clear the tasks the week they land, not the week a campaign stalls waiting on them.
How do you structure a first campaign?
Most beginner accounts fail the same way. One campaign, one ad group, forty keywords, one ad. Google then has no idea which of those forty phrases the ad was written to answer, so relevance drops, Quality Score sags, and the cost per click rises to compensate. The account is not unlucky. It is built wrong.
Build it the other way round. One campaign per budget and geography you want to control separately. One ad group per tight theme, five to twenty closely related keywords at most, with ads written for that theme and a landing page that matches. A plumber does not put “water heater replacement” and “drain cleaning” in the same ad group. Different searches, different ads, different pages.
Which campaign type
| Type | Where it runs | Best fit |
|---|---|---|
| Search | Google results for phrases people type | Direct leads and sales from demand that already exists |
| Performance Max | All Google inventory from one campaign, placed by Google’s automation | Accounts with clean conversion tracking and enough volume to learn from |
| Display | Image ads across partner sites and apps | Remarketing, and awareness at low cost per impression |
| Video | YouTube and video partners | Showing something that is hard to describe in text |
For a first campaign at a local service business, Search. Everything else is a second step.
Picking keywords and match types
Google Keyword Planner sits inside the Ads interface and gives you volumes and cost estimates for any phrase. Start from how a customer describes the problem, not how you describe the service. People search “furnace blowing cold air” far more than they search “residential HVAC diagnostics”.
Match type controls how far Google may wander from what you typed. Exact match covers your term and its close variants. Phrase match covers searches that contain your meaning. Broad match lets Google decide what counts as related, which on a small budget means paying to learn things you did not want to know. I open new accounts on phrase and exact only, then loosen once the search terms report has shown me what people actually type.
Why two advertisers pay different prices for the same click
Bid the same amount as a competitor and you can still pay more than they do. Quality Score is why. It is a 1 to 10 diagnostic Google reports for every keyword, built from three parts: expected click-through rate, ad relevance, and landing page experience. Google is explicit that the 1 to 10 number is not itself an input in the auction. What it does is surface the ad quality signals that are, so a keyword scoring badly is telling you where the extra cost is coming from.
The three columns are a diagnosis, not a grade. “Below Average” expected CTR means the copy is not answering the search, or the keyword does not match the intent behind it. Below average ad relevance usually means the phrase is not in the headline. Below average landing page experience means the page is slow, awkward on a phone, or promising something the ad did not. Whenever an account’s costs drift up for no obvious reason, I read those three columns at the keyword level first, and the answer is usually sitting there.
Writing ads Google can actually work with
Expanded Text Ads were retired in June 2022, so responsive search ads are the only text ad format you can create in a standard Search campaign. You supply up to 15 headlines and 4 descriptions and Google assembles combinations per search.
Write headlines that say genuinely different things. Main benefit. Price or offer. Service area. Licensing or years in business. Brand name. Fifteen restatements of one claim gives the system nothing to test, and you have wasted the only real advantage of the format. At least one headline needs the keyword in it or ad relevance suffers.
Pinning locks a headline into position 1, 2 or 3. That is worth doing for a legal disclaimer or a brand name you cannot let Google drop. Pin much more than that and you have hand-rebuilt a static text ad while still paying for a machine learning system you switched off.
Performance Max, and when it earns its place
Over 1 million advertisers now run Performance Max, per Google’s ads blog. One campaign reaches everything: search, Display, YouTube, Discover, Gmail, Maps. You hand over asset groups of headlines, descriptions, images and video plus a goal, and Google decides placement, audience and bid. It replaced Smart Shopping and Local campaigns.
Lebesgue’s analysis of over 5,000 ecommerce businesses found an average ROAS of 125% and an average customer acquisition cost of $17.08, varying heavily by industry and by how well the thing was set up. Google also reported that the batch of automated quality improvements it shipped during 2024 raised conversions by more than 10% on their own, which is a useful reminder that the campaign type keeps changing underneath you between the day you launch and the day you next review it.
The condition for using it is conversion tracking that works. Performance Max learns from conversion data and nothing else, so feed it garbage and it will optimize toward garbage with real efficiency. I have inherited accounts reporting hundreds of Performance Max conversions that turned out to be three form events counted several times each. PMax in 2026 is a different animal covers what has changed most recently.
How much should you budget?
The category average is close to useless for planning, because your category is not the average. WordStream’s 2025 Google Ads benchmarks put the all-industry average CPC at $5.26. Here is what the same market looks like at the level you actually pay, from Google’s keyword planning data for the United States, as of August 3, 2026:
| Search term | Cost per click | Monthly searches |
|---|---|---|
| personal injury lawyer | $163.25 | 201,000 |
| seo services | $48.59 | 74,000 |
| roof repair | $37.22 | 74,000 |
| plumber near me | $36.62 | 823,000 |
| hvac repair | $32.44 | 60,500 |
| insulation contractor | $18.59 | 33,100 |
| office space for rent | $13.37 | 18,100 |
| real estate agent | $10.63 | 201,000 |
| mobile welding | $5.62 | 6,600 |
| wedding planner | $4.28 | 450,000 |
Work the arithmetic forward from whichever row resembles your business. An insulation contractor paying $18.59 a click, converting 5% of landing page visits into enquiries and closing one in three of those, pays about $1,116 to win a customer. A mobile welding shop at $5.62 a click with the same rates pays about $337. A daily budget of $50 buys the welder roughly nine clicks a day and the insulation contractor fewer than three, which is the difference between a campaign that gathers usable data in a week and one that takes a month.
So set the daily budget high enough that the campaign collects enough clicks to learn from, or do not start it yet. Two clicks a day will never produce a signal you can trust, and you will spend three months reading noise. When the honest number is more than a business wants to spend, the fix is to shrink the box rather than the budget: fewer cities, fewer keywords, tighter hours, until the money buys real volume inside a smaller target. That is the trade I make most often when planning and monitoring a Google Ads budget for a business with a hard ceiling on spend.
Manual bids or automated bids
Setting bids by hand gives you exact control over what each keyword costs, and it earns its keep when there is too little conversion data for Google’s models to learn from. That is the honest case for it, and it is a narrow one.
Once the conversions are coming in steadily, hand it over. Smart Bidding adjusts every bid at auction time on signals no person can see or act on fast enough: device, location, time of day, browser, remarketing list membership, and Google’s own estimate of how likely that particular searcher is to convert.
| Strategy | What it does | Use it when |
|---|---|---|
| Target CPA | Chases as many conversions as it can at the cost per conversion you name | You know what a lead is worth to you |
| Maximize Conversions | Spends the budget on whatever produces the most conversions | The budget is fixed and the CPA target is not yet known |
| Target ROAS | Bids up where it predicts higher revenue per click | Conversions carry different values, as in ecommerce |
One warning attaches to all of them. Every automated strategy optimizes toward the conversion actions you defined, which means it inherits whatever is wrong with your tracking, then scales it.
How do you set up conversion tracking?
Skip this step and you are looking at clicks and guessing, and so is Google’s automation. It is the most commonly skipped part of a build and the reason plenty of accounts never improve no matter how long they run.
The Google tag (gtag.js) goes on every page of the site. That one base install lets you track any number of conversion actions without pasting a separate snippet for each. Then define the actions that represent money: form submissions, phone calls, bookings, purchases, quote requests. Not pageviews. Not scroll depth. And do not file newsletter signups next to sales, because a strategy told to maximize conversions will cheerfully buy you a thousand newsletter signups and report a triumph.
Enhanced conversions sends hashed first-party data, usually the email address somebody typed into your form, so Google can match a conversion back to its click after cookies are cleared or the person moves to another device. It does not create conversions. It recovers ones you already earned and were not being credited for.
Attribution decides which ad gets the credit when somebody clicked three of yours before converting. Google’s default is now data-driven attribution, which splits credit using the conversion paths in your own account. Last click is the only other model still supported. Google retired first click, linear, time decay and position-based, and moved every conversion action that used them onto data-driven. Data-driven is the right default for almost everybody, and the reason to change it is a specific measurement question, not a preference.
How do you tell whether it is working?
Two numbers do most of the diagnostic work, and they sit in adjacent columns.
Click-through rate tells you whether the ad matches the search. Search Engine Journal’s analysis of the last twelve months of benchmark data puts the all-industry search average at 6.64%, with Arts and Entertainment at 12.75%. Sitting well under the average on decent impression volume points at the copy, not the bid.
Conversion rate tells you whether the page delivered what the ad promised. The LocaliQ 2026 Search Advertising Benchmarks report gives an all-industry average of 8.18%, and 15.51% for automotive repair. Strong CTR with a weak conversion rate is a landing page problem, and there is no bid adjustment that fixes a landing page.
Test one variable at a time. Two ads in the same ad group, different headlines, everything else held still, left alone until each has collected enough clicks that the gap between them means something. Keep the winner, change the next thing. Headlines and calls to action are where the easy gains are, because the headline is the part that gets read.
Which audiences can you layer on?
Keywords tell you what somebody wants. Audiences tell you something about who they are, and layering the second onto the first is where local accounts tend to pick up efficiency.
In-market segments are people Google has identified as actively researching a category right now, from recent search and browsing behavior. Closest thing to intent outside a keyword. Affinity segments are the broad long-term version of the same idea, better for awareness than for leads, and custom segments let you build your own from keywords, URLs and apps your ideal customer would plausibly use. Google now calls all of these audience segments rather than audience types, so older guides describe the same features under the old name.
Two more are worth setting up on day one. Customer Match uploads your own customer list, which is as useful for excluding people who already bought as for chasing people who might. Optimized targeting, which replaced similar audiences, starts from your segments and goes looking for new people who resemble the ones who convert.
The setting that trips people up is observation versus targeting. Observation leaves your reach exactly as it was and simply reports how each audience performed, so you can move bids on evidence. Targeting restricts delivery to those audiences only. Add every audience in observation first. Switch to targeting when a month of data tells you to, never on day one.
What quietly drains the budget?
Most of the money I find wasted in accounts I take over goes the same few ways, in roughly this order.
Negative keywords
The most expensive omission in a young account. With no negatives in place, phrase and broad keywords quietly collect searches you would never have bid on deliberately, and you pay for every one of them.
Negative match types behave almost like positive ones, with one difference worth knowing. A broad match negative blocks a search only when it contains all of your negative terms, in any order. A phrase match negative blocks searches containing that exact phrase in that order. An exact match negative blocks only that precise term and nothing else. Add “free” as a broad negative and your ad stops showing for “free shoes” and “shoes for free”, but it may still show for “freedom shoes”, because that is a different word entirely.
Read the search terms report weekly, and daily through the first two weeks of a new campaign. It is the only place you see what people actually typed rather than what you bid on. I keep two layers of exclusions: account-level negative lists for terms that should never trigger an ad anywhere, such as competitor names, “jobs”, “salary”, “diy” and “free”, and campaign-level negatives for the exclusions specific to one campaign’s theme. The account-level list saves the most money over time, because it keeps working on every campaign you launch afterward without anybody remembering to apply it.
Mobile
It stopped being a segment some years ago. DemandSage’s 2026 Google Ads statistics report puts mobile at 61.9% of all Google Ads clicks, and Statista’s global traffic data has mobile at 51.48% of all global web traffic in the second quarter of 2026. A landing page that is awkward on a phone is awkward for three out of every five people you paid to send there.
Three controls are worth knowing. Mobile bid adjustments run from -100%, which switches mobile off entirely, up to +900%. If your own data shows mobile converting at half the desktop rate, a -50% adjustment pays proportionally less for those clicks instead of walking away from them. Ad scheduling, or dayparting, restricts delivery to the hours that actually convert, and I pull the device report by hour of day before setting it, because the pattern for a plumber looks nothing like the pattern for a B2B service. Call assets let somebody tap to call straight from the ad, and those calls report as conversions with duration attached.
Page speed is the easy one to ignore and the expensive one to leave alone. Google’s published threshold is a Largest Contentful Paint inside 2.5 seconds, PageSpeed Insights will tell you where you stand, and a slow page bills you twice: once in visitors who bounce before the page paints, again in a landing page experience score that pushes your cost per click up.
Assets, the things that used to be called extensions
Google renamed extensions to assets and now assembles the most relevant combination for each search rather than showing everything you supplied. Every asset that qualifies is more of the results page occupied by you and less of it available to the advertiser below you, which makes this the rare improvement that costs nothing to make.
| Asset | What it adds to the ad |
|---|---|
| Sitelink | Extra links to specific pages on your site |
| Call | A phone number or tap-to-call button |
| Location | Address, phone number and a map marker |
| Callout | Short selling points that are not clickable |
| Structured snippet | A labeled list, such as the services you offer |
| Price | Prices for services or products |
| Promotion | A sale or offer with start and end dates |
| Image | A photo alongside the text ad |
| Lead form | Captures the lead in the ad, with no site visit |
Add every type that genuinely applies to the business. Google reports performance per asset, so the weak ones are visible and worth replacing rather than guessing about.
What is actually changing in Google Ads for 2026
Three shifts matter, and they are not equally urgent. AI-generated answers now sit above the traditional results on a growing share of searches, which compresses the space where text ads compete and raises the value of each click you do win. Automated campaign types keep absorbing controls that used to be manual, which genuinely changes how the work gets done rather than just how it gets marketed. And video keeps growing, with YouTube the main channel for it.
Voice is the trend most often oversold. An estimated 153.5 million Americans used voice assistants in 2025, but the practical implication for an ad account is narrow: include the longer question-shaped phrasings people actually speak, and work the negative keyword list harder, because spoken queries are messier and waste more budget. That is keyword hygiene, not a strategy shift.
Local is the shift that matters most to a service business. Geographic targeting, location assets and Local Services Ads put you in front of searches that already carry both intent and a zip code, which is why tightly geo-targeted campaigns usually convert better than broad ones at the same cost per click. Local ads for lead generation and Google Local Service Ads ranking factors cover the mechanics.
One calculation decides whether any of this pays. Cost per click, divided by conversion rate, divided by close rate. Run it before you fund anything, then run it again on real numbers after thirty days. Nearly every account I have been handed to rescue failed that calculation on the day it launched, and nobody had done the division.
