CPC is up, results are flat. How to tell whether it’s the market, Performance Max, or your account

Updated: 29 July 2026

Short answer: it’s usually all three at once, and they can be separated

If your Google Ads cost per click has been climbing for months while conversions stay flat, you are looking at one of three situations: auction-wide inflation (happening outside your account), a budget shift into Performance Max (happening inside your account, but without your decision), or degraded conversion data (happening in the measurement layer, and routinely misdiagnosed as one of the first two).

Separating them takes about 90 minutes and three comparisons: your CPC against a market benchmark, your Search-versus-PMax spend split month over month, and your recorded conversions against actual transactions in your commercial system. Below is the diagnostic tree, the benchmarks to compare against, and an honest list of the things you cannot fix inside the account because they are market changes.

The 2026 baseline: cross-industry average Search CPC rose 12% year over year to $2.96 in Q1 2026, up from $2.64 in Q1 2025, the steepest annual increase since 2021. This is not seasonality but a shift in the baseline.

Two caveats on that figure, without which the diagnosis comes out crooked. First, published 2026 average CPCs range from $2.96 to $5.42 depending on platform scope, period and measure of centre, so the direction matters rather than the value. Second, 2026 was calmer on an annual view than the industry narrative suggests: cost per lead fell that year for the first time since before 2020. So if your costs are rising markedly faster than the market, the cause is local and findable.

What auction inflation is, and why it accelerated in 2026

The Google Ads auction resolves in a fraction of a second between advertisers bidding on the same queries. Prices rise when demand for the same inventory rises. In 2026, demand is rising for three independent reasons.

First: less free organic traffic. AI Overviews answer the question inside the results page. Companies that used to acquire informational traffic for free now enter the paid auction to get it. Digital Applied’s benchmark analysis names the mechanism directly: contributing factors are increased advertiser demand from new AI-era entrants, Google AI Overviews reducing organic CTR and pushing more budget to paid, and Performance Max creating more competition for high-value inventory. The other half of that mechanism needs stating, because without it the conclusion misleads: presence inside the AI answer itself absorbs part of the pressure. As one compilation of AI Overview and paid cost data puts it, AI Overviews reduce available attention on many research-heavy queries, while brand citation inside the overview can partially offset that pressure. A fall in organic clicks is therefore not an argument for moving budget into paid channels, but for competing in both layers at once.

Second: ads now sit inside the AI answer layer. This changes the arithmetic of a campaign. Seer Interactive tracked paid search performance across thousands of queries between June 2024 and September 2025 and found that for queries where an AI Overview was present, paid CTR dropped from 19.70% to 6.34%, a collapse of more than two thirds, while queries without AI Overviews declined far less severely. You pay the same rate and get fewer clicks from the same impression volume.

Third: automated bidding is now the default, not a choice. According to 2026 benchmark data, average Search conversion rate rose from 4.17% to 4.40%, directly correlated with Smart Bidding expanding from 68% to 78% of spend. Learning systems bid harder on high-intent signals, which raises prices exactly where you want to be present.

What a click should cost

Published 2026 figures diverge, and the divergence itself is useful information.

Three caveats before you use any of these numbers in an argument:

• The figures do not agree with each other. A $2.96 average and a $5.26 average describe different samples, not different truths.
• Almost all publicly circulating 2026 benchmarks trace back to a small number of primary datasets (WordStream/LocaliQ, Adthena, Tinuiti), republished by aggregators. Verify at the primary source before citing.
• A benchmark is a magnitude test, not a forecast for your account. Two advertisers in the same vertical routinely pay double each other’s rate depending on account quality, ad relevance and auction history.

The diagnostic tree: four questions, in order

Work from the cheapest test to the most expensive. Each step either closes a branch or moves you to the next.

Step 1. Did clicks fall, or did the price per click rise

Compare the last 90 days with the same period a year earlier, brand and non-brand campaigns separately. Four numbers matter: impressions, clicks, CTR, average CPC.

• Impressions steady, CTR falling, CPC rising is the AI Overview signature. Your ad is served, but lower and less often clicked. Check which queries trigger an overview by running them manually from a clean session.
• Impressions falling, CTR unchanged points to lost impression share, meaning price competition. Check impression share lost to rank versus lost to budget.
• CPC rising on brand terms only is usually a competitor bidding on your brand, not a market shift.

Step 2. How much budget moved into Performance Max, and did you know

This is the cause most often misattributed to the market. Performance Max decides on its own how to split budget across Search, Shopping, YouTube, Display, Discover, Gmail and Maps. If the mix shifted toward lower-intent surfaces, acquisition cost rises even with a stable CPC.

Until early 2026 this was a genuine black box. Since then Google has released visibility that you should be using:

• Channel-level reporting. The January 2026 rollout extended to all Customer IDs in both Google Ads and Search Ads 360, letting advertisers evaluate whether specific inventory contributes meaningfully to conversion goals or represents wasted spend.
• Channel contribution timeline. Since April 2026 there is a view showing how Search, YouTube, Display, Discover, Gmail and Maps each contributed over a selected period, instead of a single blended ROAS number.
• Campaign-level negative keywords. Full negative keyword support at campaign and account level, finally giving advertisers control over unwanted search queries.
• Search Partner placement visibility. From February 2026, Search Partner placements appear in the “when and where ads showed” reports.

What you still will not get, and this matters for an honest assessment: Performance Max does not provide search term reports showing the user queries that triggered ads, whether on Google.com or Search Partner Network. This remains a significant transparency limitation. Anyone promising full query transparency in PMax is promising something the platform does not expose.

Decisive test: plot PMax share of total spend against account-level CPA, month by month, over 12 months. If both curves rise together, you have your answer.

Step 3. Is your conversion data complete

This is the step most often skipped, and the one that can explain the entire increase in acquisition cost. Smart Bidding learns from conversion signals. Incomplete consent implementation and lost events mean the algorithm optimises against an impoverished picture and bids up the wrong users. The result: CPC can stay flat while real CPA climbs.

Three quick checks:

1. Compare conversions reported in Google Ads against transactions in your commercial system for the same month. A gap above 15% needs explaining.
2. Check whether consent mode is implemented in basic or advanced form, and whether conversion modelling is active at all.
3. Check whether offline or CRM conversions return to the account at all. If you sell on a lead model but only form submissions reach Ads, the algorithm is optimising for form volume, not revenue.

If the measurement layer leaks, no bid strategy change will compensate. This is the point where a Google Analytics and GTM audit delivers more than another round of bidding experiments.

Step 4. Is CPC actually the problem

The last question is the uncomfortable one. Rising CPC at stable ROAS is not a problem. It is the cost of operating in a more expensive auction. The problem is rising CPC at falling ROAS.

Calculate it in one operation: campaign revenue divided by spend, across 12 consecutive months. If the line is flat while CPC rises, traffic quality is rising alongside price. That is the normal behaviour of an auction where automation selects users better.

What you cannot fix inside the account

An honest diagnosis includes the list of things outside your control. Without it, the conversation with the board turns into a search for someone to blame.

• AI Overview pressure on comparison queries. Adthena’s analysis covering six industries from late December 2025 to January 2026, hundreds of thousands of advertisers and more than 5 million ads found the impact is highly uneven. In Telecom, Technology and Retail, AI Overviews are dominated by comparison content, which satisfies the research phase and can stop the user from clicking an ad. In Technology, queries featuring an AI Overview consistently show higher CPCs than those without, a clear signal that overview presence is pushing up the cost of visibility. In other verticals the effect runs the other way: Healthcare at 74% news themes and Financial Services at 54% FAQ themes act as intent filters, potentially protecting ad spend by satisfying low-intent users before they reach a paid link.
• The direction of the platform. Bid automation is no longer a specialist’s choice. It is the system default.
• New entrants who used to live on organic traffic. This is structural, driven by zero-click behaviour. The other half of the same problem, recovering organic and AI-answer visibility, belongs to AI Search optimisation and cannot be solved by raising the ad budget.

90-minute diagnostic checklist

1. Compare 90 days year over year, brand and non-brand separately: impressions, clicks, CTR, CPC.
2. Manually check your 10 most important non-brand queries for AI Overview presence.
3. Pull PMax share of total spend, month by month, across 12 months.
4. Open the PMax channel report and check how much budget left Search and Shopping.
5. Reconcile Google Ads conversions against transactions in your commercial system for the same month.
6. Check consent mode implementation status and whether conversion modelling is active.
7. Review lost impression share, split by rank and by budget.
8. Calculate ROAS month by month across 12 months and overlay it on the CPC curve.
9. Compare your CPC against published industry ranges as a magnitude test only.
10. Write down which findings are fixable in the account and which are market change.

Steps 1 to 4 typically deliver around 80% of the answer. Steps 5 and 6 explain most cases where account numbers and company numbers disagree.

Frequently asked questions

Is a low-double-digit annual CPC increase a lot?
It is roughly market pace. The 12% year-over-year rise in Q1 2026 was the steepest annual increase since 2021, and a 12% increase means a $10,000 monthly budget now buys roughly 10.7% fewer clicks than it did 12 months ago. If your increase is materially higher, the cause is local.

Can I go back to classic Search campaigns instead of Performance Max?
You can, but it rarely solves the cost problem on its own. The better first move is to use the controls added in 2026: campaign-level negatives, the channel report, and placement exclusions. The decision to leave PMax should come from channel report data, not from instinct.

How do I know whether my queries trigger AI Overviews at all?
The simplest test is to check 10 to 20 priority non-brand queries manually in a private browsing session, separating informational from transactional intent. Overviews appear unevenly, so the test is run on a list of queries, never a single one.

Is a lower CPC always better?
No. A lower CPC on lower-quality traffic raises acquisition cost. The decision metric is CPA or ROAS. CPC is a diagnostic indicator, not a target.

Who should run this diagnosis?
Steps 1 to 4 can be run by any PPC specialist with account access. Steps 5 and 6 require access to the analytics layer and the commercial system, so they are usually joint work between PPC and analytics.

Sources

• Digital Applied, Google Ads Benchmarks 2026: CPC, CTR, CVR by Industry, April 2026 → https://www.digitalapplied.com/blog/google-ads-benchmarks-2026-cpc-ctr-cvr-industry
• ClickCease, Google AI Overviews Are Squeezing Your Ad Budget (citing Seer Interactive, WordStream Q1 2026, Adthena/Search Engine Land), June 2026 → https://www.clickcease.com/blog/google-ai-overviews-are-squeezing-your-ad-budget-and-click-fraud-is-making-it-worse/
• Demand Local, 2026 AI Overview Presence and Paid Search CPC Statistics, May 2026 → https://www.demandlocal.com/blog/presence-and-paid-search-cpc-statistics/
• Adthena, AIOs are impacting paid search, March 2026 → https://www.adthena.com/resources/blog/aios-on-paid-search/
• Search Engine Land, What industry data reveals about the impact of Google’s AI Overviews on paid search, February 2026 → https://searchengineland.com/what-industry-data-reveals-about-the-impact-of-googles-ai-overviews-on-paid-search-470019
• ALM Corp, Google Performance Max Now Shows Search Partner Placements, February 2026 → https://almcorp.com/blog/google-performance-max-search-partner-placements-visibility-2026/
• Yellow Jack Media, Performance Max Channel Timeline, April 2026 → https://www.yellowjackmedia.com/googles-new-performance-max-channel-timeline-real-visibility-into-where-your-budget-goes/
• Benly, Performance Max 2026 Updates: New Controls & Features, March 2026 → https://benly.ai/learn/google-ads/pmax-2026-updates
• BizIQ, PPC Statistics 2026: Google Ads Benchmarks, ROI & Paid Search Data, June 2026 → https://biziq.com/blog/ppc-statistics/
• Semgence, Google Ads costs 2026 (citing LocaliQ/WordStream, February 2026) → https://www.semgence.pl/ile-kosztuje-google-ads/

Data currency note. Cost benchmarks and Google Ads platform features change faster than annually. Every figure here reflects the state of play in July 2026 and requires quarterly review. Several widely circulated 2026 benchmarks are republished by aggregators citing a small number of primary datasets, so verify at the original source before quoting them in a board document.