"Buying keywords on Google" is a common phrase, and a poor description of what happens. You do not buy a word, and nobody sells you one exclusively. You open an advertising account, tell Google which searches you want to appear on, and enter an auction that is re-run every time someone types a query. You pay per click, not per impression.
That mechanism — paid search, or search engine advertising — absorbs a large share of online advertising budgets. One measured example: in France, search took 41% of the digital advertising market in the first half of 2026, or €2.74bn, up 12% year on year (36th Observatoire de l'e-pub, SRI / UDECAM / Oliver Wyman, published 9 July 2026; net revenue booked by sales houses in France). Those are French market figures, not UK ones.
Here is how the system works, what it costs, and how to decide whether it belongs in your budget.
What you are actually buying
When you create a campaign on the search network, you declare keywords — or, increasingly, you let the algorithm infer the relevant queries from your landing page. At each search, Google runs an instant auction between eligible advertisers and decides who appears, in what order, and at what price.
Two practical consequences follow. First, there is no fixed "price of a keyword": the cost per click moves with the time of day, the device, the location and whoever happens to be bidding at that moment. Second, the maximum bid you enter is a ceiling, not an invoice: Google's documentation is explicit that what you pay is often lower. This is a different economy from organic search, which targets the unpaid results on the same page.
| Criterion | Paid search (PPC) | Organic search (SEO) |
|---|---|---|
| Time to first results | Hours to days | Months |
| Cost model | Billed per click, continuously | Cost of producing and maintaining content |
| Effect when you stop | Traffic stops the same day | Traffic declines slowly |
| Control over the message | Total, over both ad copy and landing page | Indirect; the engine decides how you are shown |
| Typical use | Launches, seasonality, testing an offer | Baseline acquisition, durable visibility |
Buying search ads also remains a minority activity among small businesses. Afnic's 2025 study of close to 2,500 French micro-businesses and SMEs, run between September 2024 and August 2025, found that 8% invested in sponsored links on search engines and 71% ran no online advertising at all (summary published by France Num on 18 March 2026; self-reported figures). A French sample again, but a reminder that this market is less crowded at the small-business end than the noise suggests.
Where to buy: Google, and possibly Microsoft
Google Ads is the obvious entry point in most markets. Before assuming it is the only one, check your own country in a public dashboard such as StatCounter Global Stats, read as an order of magnitude: it measures page views across a panel of websites, not a panel of people.
The second platform is Microsoft Advertising, formerly Bing Ads, which serves Bing and its partner network. Its own home page states that there is no minimum spend and that billing is per click. The audience is much smaller, so expect lower volume. You will often read that clicks are cheaper there; plausible in some sectors, but verify it campaign by campaign rather than assume it.
On budget, Google Ads asks for no minimum. You set an average daily budget per campaign. The official rule is worth memorising: spend on a single day can reach twice that budget, but the monthly bill will not exceed 30.4 times it (Google Ads Help, average daily budgets). A budget set at £20 a day therefore caps the month at roughly £608.
How your position is decided
Where your ad appears depends on Ad Rank. Google documents six components: your bid, the quality of the ad and the landing page, the ad rank thresholds, the competitiveness of the auction, the context of the search (terms, device, location, time) and the expected impact of your assets and ad formats (Google Ads Help, Ad Rank).
One point deserves correcting, because the opposite circulates widely. The 1-to-10 Quality Score shown at keyword level is not a multiplier applied to your bid. Google states plainly that it is not used in the ad auction, and that it is a diagnostic tool calculated over the last 90 days from expected click-through rate, ad relevance and landing page experience (Google Ads Help, Quality Score). Quality does count in the auction — but it is assessed in real time, not through that score.
The levers you actually control
- Consistency between the query, the ad copy and the landing page. The most profitable lever, and the most frequently neglected.
- Landing page quality: content that delivers on the ad's promise, a clear path, a short form. A UX audit often fixes more than raising a bid does.
- Loading speed and mobile rendering, which weigh on measured experience and on conversion rate. See website speed optimisation.
- Exclusions: negative keywords, geographic areas, schedules. This is what stops budget draining into clicks you never wanted.
What changed recently: AI now picks part of the targeting
The old model — "I list my keywords, I control everything" — is no longer the default. Since 2025, Google has been rolling out AI Max for search campaigns: a layer that broadens query matching, rewrites headlines and descriptions, and can send the click to a landing page other than the one you declared. The documentation confirms that AI Max is on by default in new search campaigns, and that turning it off means disabling each of its settings (Google Ads Help, set up AI Max).
In practice, you are buying less a list of keywords than an intent interpreted by an algorithm. That is neither good nor bad in itself, but it moves the work: watch the search terms report, keep enriching the negative list, and check that the pages chosen automatically are the ones you wanted shown. On a small budget, start with final URL expansion switched off.
A first campaign, step by step
- Define a measurable conversion before you spend anything: a call, a form, a quote request, an order. Without it, no later decision is possible.
- Pick 10 to 30 queries close to the point of purchase, not informational ones. "Emergency boiler repair Leeds" beats "boilers".
- Create one campaign per offer, and a dedicated landing page per ad group.
- Write at least two ads per group, reusing the wording of the query, and stating price or service area when that is how customers filter suppliers.
- Set location targeting tightly, and start with a negative keyword list on day one ("free", "jobs", "definition", competitor names you would rather not bid on).
- Let it run two to four weeks before judging. Killing a campaign after three days proves nothing.
Measure, or do not buy
This is the entry condition, and where a lot of companies quietly lose money. In the same 2025 Afnic study of French small businesses, 61% said they did not really monitor the performance of their online acquisition activity. Buying clicks without tracking what they produce is funding visibility without knowing whether it pays.
The minimum is conversion tracking configured correctly, a reconciliation with your actual sales, and a dashboard someone reads every month. Looker Studio, formerly Data Studio, is free and enough to bring Google Ads, Microsoft Advertising and your analytics into one view. The indicator that matters is not cost per click but cost per acquired customer, set against your margin — the purpose of our traffic management and measurement work.
What I have learned running these accounts
I took Google's and Bing's advertising certifications when I started running campaigns for clients. Ten years on, what stays with me has nothing to do with the badges. The campaigns that work for small and medium businesses are almost always the narrowest: few queries, all close to purchase, and a landing page that says exactly what the ad said. The failures I have seen rarely came from bidding; they came from targeting that was too broad, or a page that did not convert.
There is also a second return that rarely appears in the reporting. Within a month, an account tells you which wordings trigger enquiries and which only attract browsers — worth reusing to steer your organic search content, where the same lesson might have taken a year to surface.
Common questions
Can you really buy a keyword on Google?
Not in the sense of an exclusive purchase. You take part in an auction that is re-run at every search, and you pay only when someone clicks your ad. Several advertisers bid on the same keyword at once, and the order they appear in can change from one query to the next.
What minimum budget do you need to start on Google Ads?
Google sets no minimum: you choose an average daily budget per campaign. A single day's spend can reach twice that budget, while the month stays capped at 30.4 times the daily figure. The real constraint is different: a budget too small to generate clicks produces data you cannot act on.
Should you advertise on Bing as well?
Microsoft Advertising can be worth a limited-budget test, particularly in B2B, but volumes are much lower than Google's in most markets. Check your own country's search share first, and stabilise Google Ads before opening a second account, so you have a benchmark.
Does a high Quality Score lower my cost per click?
The 1-to-10 score shown in the interface is a diagnostic tool; Google states it is not used in the ad auction itself. The real quality of the ad and the landing page is assessed at auction time and does influence your rank and your price. Improving page relevance is worthwhile; the displayed number is not the lever.
How long before you know whether a campaign works?
Allow two to four weeks of continuous delivery, provided conversion tracking was in place on day one. Below that, click volume is too low to separate signal from noise.
Should you leave AI Max switched on?
It depends on how much budget you can afford to spend while learning. Because it is on by default and can broaden matching and change the landing page, a small account often does better disabling the expansion settings first, reading the search terms report, then re-enabling them once targeting is proven.