There is no such thing as a high cost per click. There is only a cost per click that is high relative to what the click earns you — and a cost per click that is higher than it was, which is a completely different problem with completely different causes.
Almost every guide to this question skips that distinction and goes straight to a list of tips. So you end up pausing keywords that were profitable, rewriting ads that were fine, and chasing a Quality Score that Google’s own documentation says is not an input to the auction. This post is built the other way round: first work out which of three problems you actually have, then read only the causes that apply to it.
High, rising, or spiked? Three different problems
It has always been like this. You have no baseline that felt cheap.
Usually not a faultYou are comparing against a number you invented. Start at cause 1, then read why CPC is the wrong target.
The distinction matters because the diagnostics do not overlap. A CPC that has always been £6 in a market where the average is £8 is not a fault at all. A CPC that has crept from £3 to £4 over five months is almost always the market moving around you. A CPC that doubled on a Tuesday is almost always something that happened in your account on that Tuesday — and change history will name it in under a minute.
How your CPC is actually set
Two facts do most of the explaining, and neither is widely understood. The first is that your bid is a ceiling, not a price. Google is explicit about this in its documentation on actual cost-per-click: you are “often charged less — sometimes much less” than your maximum CPC, because you only pay “what’s minimally required to clear the Ad Rank thresholds and beat the Ad Rank of the competitor immediately below you.”
The second follows from it, and it is the single most useful idea in this post: your CPC is set by the advertiser below you, not by you. Which means your CPC can rise sharply while your bid, your quality, your ads and your position all stay exactly where they were.
Read the second row. You bid £4.00 and pay £3.13 — because that is all it takes to stay ahead of the advertiser beneath you. Competitor A bid £6.00 for the position above and pays £5.34 to hold it. Higher quality is buying you a better position at a lower price than the advertiser above you is paying.
Press the toggle above and watch what happens. Nothing about the highlighted advertiser changes — same bid, same quality, same position — and the price goes up by roughly a quarter, purely because someone underneath got more aggressive. That is not a hypothetical. It is the most common single answer to “why is my CPC rising”, and it is invisible from inside your own account unless you go looking for it.
The Ad Rank side of the formula is where quality enters. Google’s documentation on Ad Rank lists the inputs as your bid, the quality of your ads and landing page, the Ad Rank thresholds, the competitiveness of the auction, the context of the search — location, device, time, the nature of the query — and the expected impact of your assets and ad formats. Note what that means in practice: Ad Rank is recalculated for every single auction, so your “average CPC” is an average over thousands of separate contests you never see individually.
The nine causes, by frequency
Ordered by how often each one turns out to be the real answer across real accounts, not by how interesting the mechanism is. The meter on each card tells you how commonly it is the culprit and which of the three problems it explains. If you are short on time, causes 1 to 4 account for the large majority of cases.
You are comparing against a number you invented
Clicks are not priced in the abstract. They are priced by what your competitors can afford to pay, which is a function of what a customer is worth in your category. In WordStream’s 2026 benchmark study — covering April 2025 to March 2026 — the cross-industry average search CPC was $5.42, but the spread around it is enormous:
| Industry | Average CPC |
|---|---|
| Arts & Entertainment | $1.63 |
| Restaurants & Food | $2.05 |
| Travel | $2.14 |
| All industries | $5.42 |
| Dentists & Dental Services | $8.00 |
| Home & Home Improvement | $8.33 |
| Attorneys & Legal Services | $9.87 |
A dentist paying $8 a click is paying the going rate. An arts venue paying $8 a click has a serious problem. Same number, opposite verdicts — which is why the benchmark that actually matters is not your industry’s average at all. It is your own breakeven.
Workable. Profitable with room to absorb a bad month. Worth defending, not worth panicking about.
Run your own figures through that before you touch a single setting. If the headroom number comes back comfortably above 1, your CPC is not your problem and the remaining eight causes are a distraction — go and work on volume instead. If it comes back under 1, note that the fastest lever is almost never the click price. Doubling a 2% conversion rate to 4% does exactly what halving your CPC would do, and unlike your CPC, it is entirely within your control.
Your competitors changed, not your account
This is the cause the auction diagram above demonstrates, and it is where you should look first for any gradual rise. Auction insights is the only report that shows you the other side of the auction. Segment it by month rather than reading a single date range — a new domain appearing in the list, or an existing one whose overlap rate and position above rate have climbed, is your answer.
Two readings are worth learning to tell apart. If overlap rate has grown, more advertisers are showing alongside you and the auction is simply busier. If overlap is flat but position above rate has grown, the same competitors are bidding harder or improving their quality — you are being outranked by people who were always there. The first is a market you may not want to fight. The second is often winnable on quality rather than price.
Note the report’s blind spot: Google will not show auction insights when your impression share is below 10%, so the campaigns most likely to be getting squeezed are the ones where the data goes quiet. If that is you, work from the cost side instead — average CPC by month, segmented by device and by network.
Competitive drift is slow, continuous and easy to miss between quarterly reviews, which is exactly the kind of thing worth watching automatically rather than remembering to check. Our anomaly detection baselines each account against its own history, so a CPC curve bending away from its normal shape surfaces as a message rather than as a quarter-end surprise.
Your keywords are matching wider than you think
Match types stopped being literal years ago. Exact match now covers close variants — misspellings, singular and plural forms, function words, reordering, abbreviations and stems — so [kitchen fitting] can serve against queries you never wrote down. Phrase match matches on meaning rather than word order. Broad match reaches furthest of all, and its reach changes over time as Google’s understanding of your account changes.
The consequence for CPC is indirect but large. You did not raise your bid; you started entering different auctions, and those auctions have different competitors with different economics. A plumber bidding on emergency plumber who drifts into boiler replacement cost is now competing against advertisers with a far higher customer value, and they will outbid them all day.
Layer on AI Max, which Google is rolling into more Search campaigns through 2026 — including as the upgrade path for Dynamic Search Ads. Its search term matching is designed to expand beyond your keywords, and it expands hardest on exact and phrase match, precisely because those are the narrowest. Google’s AI Max FAQ confirms negative keywords are still respected with it on — which makes your negative list the main control you have left.
The fix is unglamorous and permanent: read the search terms report weekly, sort by cost descending, and add negatives for anything that is not a buyer. Then check that the expensive terms are landing in the right ad group rather than being served by a looser keyword elsewhere in the account. This is the same discipline as the search terms section of our audit checklist, and it is the highest-yield twenty minutes in Google Ads.
Performance Max is buying your search traffic
Google’s stated rule is that a Search campaign with a keyword matching the query exactly should take priority over Performance Max. The evidence says the rule is leakier than it sounds. Optmyzr’s study of 503 accounts found keyword overlap between Search and PMax in 91.45% of them, across all match types including exact — and when the two did overlap, Search was the better performer far more often than PMax, winning on click-through rate in 28.37% of cases against PMax’s 15.98%.
There are two distinct CPC effects and they get conflated. The first: PMax spreads across cheaper inventory, so it reports a lower CPC than Search while taking the same queries — your blended number improves while your Search campaigns get worse, which is why this hides so well in an account-level view. The second: when PMax and your own Search campaign are eligible for the same auction, you are adding pressure to a market you were already competing in.
Efficiency in the channel has been moving the wrong way, too. Channable’s 2026 benchmark, drawn from €1.38 billion of verified spend across more than 10,000 European advertisers between June 2025 and June 2026, reported a 15% rise in CPC on Shopping and Performance Max alongside a 46% year-on-year fall in PMax return on ad spend.
Practical controls, in order of effect: add campaign-level negative keywords to the PMax campaign — the limit went from 100 to 10,000 in 2025, so there is no longer an excuse for a thin list; use brand exclusions to stop PMax absorbing brand searches you would have won organically; and check the PMax search terms view against your Search campaigns rather than assuming the priority rule is holding. This pattern is close to universal in ecommerce accounts, where PMax carries the majority of spend.
Your bid strategy has no ceiling
If your CPC spiked rather than crept, start here, and start with change history. It records who changed what and when, and it resolves most spikes in a minute flat.
The mechanism people miss: Smart Bidding has no maximum CPC. On manual bidding, your max CPC is a hard cap. Switch to Maximize conversions, Target CPA or Target ROAS and that cap disappears — the strategy will bid whatever an individual auction seems to be worth, which on a high-intent query can be many multiples of what you would ever have typed in. That is frequently correct behaviour and still a shock when you first see it. If you need a ceiling back, bid limits exist but only on portfolio strategies, not standard ones, and Google discourages them because they constrain the optimisation you switched to automation for.
Three specific traps worth checking by name:
- Maximize clicks with no bid cap. It is doing exactly what you asked — buying the most clicks your budget allows — and it will happily pay far above your old manual bids to do it. Almost always the wrong strategy if you have conversion data.
- Enhanced CPC leftovers. Google retired eCPC for Search and Display in the week of 31 March 2025 and migrated the campaigns that had not been moved to Manual CPC. If a campaign has been quietly sitting on manual bidding ever since, its bids are frozen at whatever they were and its performance has been drifting away from the market for over a year.
- A target that moved. Raising a Target CPA or loosening a Target ROAS tells the system it may pay more per conversion, and it will. This is the correct trade if you wanted volume — just make sure it was a decision rather than a default.
Whatever you change, change one thing and wait. Significant bid strategy edits restart the learning period, and stacking three changes on top of each other guarantees you never learn which one mattered — the same sequencing rule that governs budget changes.
Ad quality is dragging your Ad Rank down
Here is where nearly every article on this topic goes wrong, so it is worth being precise. Ad quality genuinely does affect what you pay: it is an input to Ad Rank, Ad Rank determines what you must clear to win, and a better-quality ad can hold a higher position at a lower price than a worse one bidding more. That much is real, and the auction diagram above shows it happening.
What is not real is the idea that the 1–10 Quality Score in your keyword table is the thing being used.
Treat it the way Google describes it — as a diagnostic. Ignore the composite number and read the three components, because each points at a different fix:
- Expected click-through rate below average. Your ad is not compelling for that query. Usually a relevance problem: the keyword and the headline are not talking about the same thing.
- Ad relevance below average. The ad group is too broad. Split it so the ad can name the thing the searcher typed — this is the actual argument for tight ad groups, and it is an Ad Rank argument rather than an aesthetic one.
- Landing page experience below average. The most-ignored and often the most valuable. Slow pages, a mismatch between ad promise and page content, or a page that makes the visitor hunt for what they were promised. Fixing it raises quality and conversion rate, which moves both sides of the breakeven calculation at once.
You are buying a position you do not need
The top of the page is priced at a steep premium, and the premium is not proportional to the return. Position one collects more clicks, but the incremental clicks are disproportionately the least considered ones — people who click the first thing they see. Position three often converts at a similar or better rate for meaningfully less money.
The usual culprit is target impression share bidding, which optimises for visibility and is close to indifferent to cost. Set to the absolute top of the page at a high target, it will pay what it takes. That is a legitimate strategy for defending brand terms against competitors bidding on your name; it is an expensive mistake on generic terms, where it buys position rather than customers.
Test it rather than debating it. Move one campaign from target impression share to Maximize conversions for a fortnight and compare cost per conversion, not CPC and not position. The result is usually decisive, and it is frequently the single largest saving available in a full account audit.
Ad Rank thresholds are pricing a thin auction
“Nobody else is bidding on this, so why does it cost so much?” is one of the most confusing things in Google Ads, and it has a clean answer. Ad Rank thresholds are reserve prices — a floor you must clear to show at all, set independently of competition. Google states they vary with your ad quality, the position on the page, user signals like location and device, and the topic and nature of the search.
When you are the only advertiser clearing the threshold, there is no competitor beneath you to price against, so you pay the reserve. And because the threshold is quality-sensitive, a weak ad on an uncontested query can genuinely cost more than a strong ad on a contested one. This is the case where improving quality is not merely the cheaper lever — it is the only lever, because no amount of extra bid buys you past a threshold you fail on quality.
It also explains a pattern that looks like a bug: sensitive or heavily regulated topics carry higher thresholds, so the same ad quality costs more in some categories than others. If this is your situation, the honest options are to improve landing page experience and ad relevance, or to accept the floor and make the maths work on conversion rate instead.
The auction floor itself moved
Sometimes it really is not you. Three separate forces push the whole floor upward, and it is worth knowing which is which before you spend another week optimising against something you cannot change.
Structural demand. More advertisers, more automation bidding into the same inventory, and AI Overviews absorbing space above the fold on informational queries. Costs in this channel have roughly doubled over a decade — WordStream’s series has the all-industry average moving from $2.32 to $5.42 since 2016.
Seasonality. Q4 is not a normal quarter. Retail auctions routinely run tens of percent above baseline through Black Friday and December, and a budget calibrated in Q1 will not stretch through Q4 at the same volume. Judge year-on-year rather than month-on-month, or you will keep diagnosing November as a fault.
Google’s own hand on the dial. This one is rarely said out loud, so here it is plainly. In the 2023 US Department of Justice antitrust trial, Jerry Dischler — then vice president of Google’s ads business — testified that Google adjusts its search ad auctions to hit internal revenue targets, with changes referred to internally as “tunings” and “knobs”, and that such adjustments could raise prices by around 5 to 10%. As reported at the time, he also told the court: “We tend not to tell advertisers about pricing changes.” Search Engine Land’s walkthrough of the trial exhibits covers the mechanics in more detail.
None of that is a reason for fatalism, and it is not an excuse to stop optimising — the eight causes above are all still worth more to you than this one. But it is a reason to hold your benchmarks loosely, to measure your performance against your own trend rather than against a remembered CPC from three years ago, and to stop looking for a fault in your account when the fault is not there.
What to do, in order
Sequencing matters as much as the individual fixes, because early causes make later ones unreadable. Ninety minutes worked top to bottom beats a month of ad-hoc changes.
- Work out your breakeven CPC (5 min). If you have headroom, stop — you have a volume problem wearing a cost problem’s clothes.
- Establish which problem you have (5 min). Pull average CPC by month for 24 months. A step change sends you to change history; a slope sends you to auction insights.
- If it spiked: read change history first (5 min). Bid strategy changes, target changes, budget changes, new campaigns. The answer is usually sitting there with a date and a name on it.
- If it is rising: check auction insights and search terms (20 min). New entrants first, then the queries you are actually buying. Causes 2, 3 and 4, in that order.
- Then fix quality, not price (ongoing). Ad relevance and landing page experience lower your effective CPC permanently, unlike a bid cut, which only removes you from auctions.
- Wait 7 to 14 days before judging anything. Then change the next single thing.
If you would rather not do this by hand every month across several accounts, this is precisely the shape of work an AI Google Ads agent is good at: read the account, notice the CPC curve bending, name which of these nine causes fits the evidence, and rank the fixes by what each one is costing you. Adszy does that read-only until you approve — nothing reaches the account without a yes. You can also point Claude or Codex at the same data through our MCP server if you would rather ask the questions yourself.
The uncomfortable part: CPC is not a KPI
Cost per click is trivially easy to reduce. Bid less, target broader, chase cheaper positions, and buy low-intent traffic in bulk. Every one of those will bring your average CPC down and most of them will make you poorer. It is one of the few metrics in the account you can improve by making worse decisions.
So the question worth asking is not “why is my CPC so high?” but “is my cost per click justified by what the click returns, and if not, which end of that ratio is easier to move?” Nine times out of ten the answer is the returns end — conversion rate, offer, landing page, and the quality of the traffic you let in — not the price of the click itself.
Both halves should surface on their own rather than wait for someone to go looking. That is the job we built Adszy to do: watch the account continuously, rank what each problem is costing you, draft the fix, and apply only what you approve. Ask it why your CPC went up last month and it answers with the evidence attached — or put the answer in your weekly report so it never becomes a surprise. The free plan covers one account with no card, or watch it work end to end in the interactive demo first.