按点击付费（PPC）：没有广告，他们还会买吗？
 - Arfadia.jpg "按点击付费（PPC）：没有广告，他们还会买吗？ - Arfadia")
Pay-per-click (PPC) is an advertising pricing model in which you're charged when someone clicks your ad, not when they buy. The click is the billable event, so a PPC invoice measures traffic you paid for, not sales the ads caused.
That gap matters more than most guides admit. When researchers ran controlled experiments on eBay's paid search, naive regression estimates put the return above 4,000%, while the experiment itself measured an ROI of -63% (VERIFIED: NBER Working Paper 20171, later published in Econometrica, 2015; US field experiment). Same ads. Different methods. Opposite conclusions.
This entry covers how PPC pricing works, where the auction fits, and the one question most PPC writing skips. Would a customer have bought anyway, without the paid click?
Pay-Per-Click Explained
Start with the name. It describes how you pay, not where your ad runs. Google's own documentation says that with cost-per-click bidding you pay for each click on your ads, and that CPC pricing is sometimes known as pay-per-click (VERIFIED: Google Ads Help).
What Actually Triggers the Charge
The charge happens at the click. It doesn't wait for a form fill, a phone call, or a purchase. A click can be relevant, well targeted, and followed by a sale, and you still won't have proof from the click alone that the ad produced that sale.
Why this matters: every metric built on top of the click inherits the same blind spot, from cost per acquisition to return on ad spend. Those numbers describe what happened after people clicked. They can't describe what would have happened if the ad had never shown, and that missing comparison is where PPC spend can be misjudged.
A Pricing Model, Not a Channel
Paid search is the best-known home for PPC, which is why the term often gets folded into search engine marketing as a channel. Click-based pricing isn't limited to the classic results page, though. It now shows up inside AI-generated answers too, covered further down.
In the US, search advertising revenue, a figure that includes AI search, reached $114.2 billion in 2025, up 11.0% and equal to 38.8% of all digital ad revenue, according to an unaudited industry association report from IAB and PwC (REPORTED; US only). Search revenue and PPC aren't the same measure. The figure shouldn't be read as the size of a PPC market, and it says nothing about what a click is worth to you.
How the Auction Prices a Click
Your bid isn't the whole story. Google says Ad Rank, which decides whether and where an ad appears, draws on six things: your bid, the quality of your ads and landing page, minimum Ad Rank thresholds, the competitiveness of the auction, the context of the person's search, and the expected impact of your assets and other ad formats (VERIFIED: Google Ads Help). That's why a higher bid can still lose, a mechanism the Ad Rank glossary entry breaks down in detail.
Quality Score sits next to Ad Rank, but it isn't the same thing. Google describes Quality Score as a diagnostic tool and states that it isn't an input in the ad auction; its three components are expected click-through rate, ad relevance, and landing page experience (VERIFIED: Google Ads Help).
PPC vs. CPC vs. SEO
Definitions follow the platforms' own documentation, accessed October 2026. Billing settings change, so check the current help page before you configure a campaign.
| Term | One-line definition | What triggers the charge | Source |
|---|---|---|---|
| Max CPC | The highest amount you set for a click; you're often charged less, sometimes much less. | A click, priced at or below this ceiling (bid adjustments can push the charge above it) | Google Ads Help, Actual CPCVERIFIED |
| Actual CPC | The final amount charged for a click: only what's minimally required to clear Ad Rank thresholds and beat the Ad Rank of the competitor immediately below you. | A click | Google Ads Help, Actual CPCVERIFIED |
| Average CPC | The total cost of your clicks divided by the total number of clicks; it can differ from your max CPC. | Not a charge; a reporting metric | Google Ads Help, Average CPCVERIFIED |
| Cost-per-click (platform metric) | Alphabet's definition: click-driven revenues divided by our total number of paid clicks. | Not a charge; a revenue metric reported by the platform | Alphabet Form 10-Q and 10-KVERIFIED |
| Target CPA | A bid strategy that aims for a cost per acquisition; conversion tracking has to be set up before you can use it. | Set by the campaign's billing option, not by the bid strategy itself | Google Ads Help, Target CPA; Pay for conversionsVERIFIED |
| Pay for conversions | A Display campaign option to pay for conversions rather than clicks or interactions; the account needs more than 100 conversions in the last 30 days. | A conversion | Google Ads Help, Pay for conversionsVERIFIED |
| oCPC (ChatGPT ads) | Conversion-optimized campaigns with click billing, chosen through a relevance-weighted, second-price auction. | A valid click, not a conversion | OpenAI Help Center, as of October 2026VERIFIED |
| Invalid traffic | Clicks and activity Google detects and filters as invalid. | No charge if detected before billing; a credit where appropriate and possible afterward | Google Ads Help, Invalid trafficVERIFIED |
These three terms get mixed up constantly. PPC names the payment arrangement. CPC does double duty, naming the same pricing method and also the cost metric you read in reports. Google's wording also treats the pricing terms as overlapping. SEO is a different discipline altogether, because it works on organic results, the unpaid listings that sit outside the ad auction.
Many teams run both. A search engine optimization program builds visibility you don't pay for per click, while PPC buys placement on demand. The two aren't independent, though. When the same person could reach you through an organic listing or a paid one, the paid click may simply be replacing a free one, and that overlap is exactly where the next section gets uncomfortable.
When a Paid Click Buys Demand You Already Had
Each row pairs a common PPC situation with what the cited experiments found. The last column is editorial analysis built on those findings, not a measured result.
| Situation | What the evidence shows | What the paid click is likely doing | Suggested next step (analysis) |
|---|---|---|---|
| Your own brand term, no competitors bidding | eBay's brand keyword ads had no measurable short-term benefit. Without competitors present, brand ads had a 1% to 4% effect.NBER w20171, VERIFIED; Simonov, Nosko, Rao, Marketing Science 2018, REPORTED | Mostly buying visits that would have arrived through organic or direct traffic | Test a pause or holdout before adding budget |
| Your own brand term, competitors bidding | Competitors took 18% to 42% of clicks from brands that didn't advertise, and the return on defensive brand ads was strongly positive.Simonov, Nosko, Rao, Marketing Science 2018 (Bing), REPORTED | Defending demand that competitors would otherwise capture | Keep the spend, and measure it against a holdout |
| Non-brand term, new or infrequent customers | New and infrequent users were positively influenced by eBay's non-brand ads.NBER w20171, VERIFIED | Adding sales that might not have happened otherwise | Prioritize these audiences and confirm the lift with a test |
| Non-brand term, frequent customers | Frequent users weren't influenced by the ads, yet they accounted for most of the advertising expense, so average returns were negative.NBER w20171, VERIFIED | Paying for purchases that were already coming | Check how much spend these users absorb before scaling |
| Any campaign judged on attributed results alone | Naive OLS estimates of eBay's return exceeded 4,000% without controls and 1,400% with controls, while the experiment measured -63% (95% CI: -124% to -3%).NBER w20171 and Econometrica 83(1), 2015, VERIFIED; US | Unknown, because attribution credits the click, not the cause | Treat attributed ROI as a ceiling, not a verdict |
| You want to measure the difference directly | Conversion Lift measures incremental conversions and reports incremental CPA, defined as total spend divided by incremental conversions. It isn't available in every account, and access runs through a Google account representative.Google Ads Help, VERIFIED | Measurable, if a lift study or holdout is in place | Ask about Conversion Lift access, or run a geographic holdout like eBay's |
This is the part of PPC that rarely makes it into a definition. A paid click can be billable, relevant, and followed by a purchase, while the purchase would have happened through an organic listing or a direct visit anyway. No standard attribution report can show you that counterfactual on its own.
What eBay Found When It Switched Ads Off
Researchers Tom Blake, Chris Nosko, and Steven Tadelis ran large-scale field experiments on eBay's paid search, described in NBER Working Paper 20171 in 2014 and later published in Econometrica, volume 83, issue 1 (2015). For brand keywords, eBay halted its search ads on Yahoo! and Microsoft (MSN) while Google kept running as the control. The result was blunt: brand keyword ads had no measurable short-term benefit (VERIFIED: NBER w20171).
Non-brand keywords told a subtler story. Ads were suspended in roughly 30% of US designated market areas, with the rest serving as controls. New and infrequent users responded to the ads, but frequent users, whose buying didn't change, accounted for most of the advertising expense, so average returns came out negative. The experimental ROI was -63%, with a 95% confidence interval running from -124% to -3%, while naive OLS estimates exceeded 4,000% without controls and still topped 1,400% with them (VERIFIED: NBER w20171; US).
Tadelis put it plainly in Berkeley Haas coverage of the study in January 2014, in the university's coverage. In his account, once the paid ads were turned off, almost all of the traffic they had carried was replaced by free channels (VERIFIED: Berkeley Haas Newsroom).
Worth checking before you generalize, though. This was one very large, very well-known advertiser. Its name was already the obvious thing to type.
The Brand Keyword Question Has Two Answers
Later experiments on Bing by Simonov, Nosko, and Rao, published in Marketing Science, volume 37, issue 2 (2018), looked at brand ads under different competitive conditions (REPORTED: journal abstract via catalog records). When no competitors were present, brand ads had a positive effect of 1% to 4%. When a brand didn't advertise and competitors did, those competitors took 18% to 42% of the brand's clicks, which made the return on defensive brand advertising strongly positive.
These findings don't cancel each other out. They answer the same question under two conditions. If nobody else bids on your name, your brand ads may be paying for clicks you'd have earned for free. If competitors are bidding on it, staying out can hand them the traffic that was looking for you in the first place.
Attributed Conversions vs. Incremental Conversions
Attribution assigns credit for a conversion to the ad that came before it. Incrementality asks whether the conversion needed the ad at all. They're different questions, and the eBay gap between more than 4,000% and -63% shows how far an observational estimate can sit from an experimental one.
Google offers one way to test the difference. Conversion Lift measures the incremental conversions driven by Google Ads campaigns, and it reports incremental CPA, defined as total spend divided by incremental conversions (VERIFIED: Google Ads Help). It isn't available for every account. Access runs through a Google account representative.
A Three-Question Test Before You Scale
Before you raise a PPC budget because the reported numbers look good, run through these three questions.
- Would they have found you anyway? If the search is your own brand name and you already appear first in the organic results, the paid click may be buying a visit that was coming regardless.
- Is anyone else bidding on that term? If competitors show up on your brand searches, the Simonov, Nosko, and Rao results point the other way, and defensive spend can earn its keep.
- Can you hold some traffic back? A holdout, whether a lift study or a geographic split like the one eBay used, is the cleanest way to see what actually disappears when the ads stop.
None of this means PPC doesn't work. It means the click is a receipt, not a verdict. Organic demand built through work such as content marketing and organic search can change how much of your paid traffic is genuinely new, which is one more reason to test rather than assume.
PPC Beyond the Search Results Page
As of October 2026, pay-per-click pricing reaches beyond the classic results page and into AI-generated answers, with rules that differ from one platform to the next.
Ads Inside AI Answers
On Google, advertisers running Performance Max, Shopping, or Search campaigns with broad match, including AI Max for Search, were made eligible to have their ads appear in AI Overviews and AI Mode, according to a May 2025 Google announcement that described ads in AI Mode as still in testing at the time (VERIFIED: Google Ads & Commerce Blog).
OpenAI's help documentation describes ChatGPT ads that work differently from keyword search. Conversion-optimized campaigns with click billing charge for valid clicks, not conversions. Ads are chosen through a relevance-weighted, second-price auction. OpenAI recommends starting CPC campaigns with a maximum bid of US$3 to US$5 per click, and the context hints advertisers provide guide matching but aren't exact-match keywords (VERIFIED: OpenAI Help Center).
Microsoft takes another route. Its documentation says ads in Copilot are created automatically from a campaign's existing assets, including text and images, and it names Performance Max as the best way to increase exposure there (VERIFIED: Microsoft Advertising Help).
For click-billed campaigns, the billing event stays the same across all three. Someone clicks. The question from the previous section doesn't go away either, because a click on an ad inside an AI answer still can't tell you whether that customer would have bought without it.
PPC Frequently Asked Questions
Is PPC the same as CPC?
Mostly, yes, as a pricing term. Google notes that CPC pricing is sometimes known as pay-per-click. CPC also names a cost metric in reports, so it's clearer to say PPC for the payment model and CPC when you mean what your clicks actually cost.
Does the highest bid always win a PPC auction?
No. On Google, Ad Rank weighs your bid alongside ad and landing page quality, thresholds, auction competitiveness, search context, and the expected impact of your assets. A lower bid backed by stronger quality can still outrank a higher one.
Why can a PPC campaign look profitable when it isn't?
Because reported conversions credit the click, not the cause. In eBay's experiments, naive estimates of return exceeded 4,000% while the controlled test measured -63%. Attributed results can look strong while the ads add very little.
Should I bid on my own brand name?
It depends on who else is bidding. eBay's brand ads showed no measurable short-term benefit, while Simonov, Nosko, and Rao found that competitors could take 18% to 42% of clicks from brands that stayed out. Check the auction before you decide.
How can I tell whether PPC adds sales I wouldn't have gotten?
Run a holdout. Google's Conversion Lift, where your account has access through a Google representative, measures incremental conversions and incremental CPA. Without a test, you're reading attribution, not impact.
Do ChatGPT ads charge per click?
As of October 2026, OpenAI offers CPC campaigns, and its conversion-optimized campaigns with click billing charge for valid clicks rather than conversions. Ads are selected through a relevance-weighted, second-price auction.
Related Terms
- Cost-Per-Click (CPC): the pricing term and cost metric that PPC overlaps with.
- Cost Per Acquisition (CPA): what a conversion costs once clicks are counted against results.
- Click-Through Rate (CTR): how often an impression turns into the click you pay for.
- Quality Score: Google's diagnostic rating of ad quality, separate from the live auction.
Sources
- Google Ads cost-per-click definition, Google Ads Help, living documentation, accessed October 2026.
- Google Ads: About Ad Rank, Google Ads Help, accessed October 2026.
- Google Ads: About Quality Score, Google Ads Help, accessed October 2026.
- Google Ads: About Conversion Lift, Google Ads Help, accessed October 2026.
- NBER Working Paper 20171, Blake, Nosko, and Tadelis, "Consumer Heterogeneity and Paid Search Effectiveness: A Large Scale Field Experiment," May 2014; published in Econometrica 83(1), 2015.
- Berkeley Haas paid search study, Berkeley Haas Newsroom, January 17, 2014.
- Simonov, Nosko, Rao (2018), "Competition and Crowd-Out for Brand Keywords in Sponsored Search," Marketing Science 37(2):200-215, catalog record and abstract.
- Alphabet Form 10-Q, Q2 2026, Alphabet Inc., filed with the US SEC, July 2026.
- IAB/PwC Internet Advertising Revenue Report, full year 2025, April 2026; data not audited by PwC.
- OpenAI: Ads in ChatGPT basics, OpenAI Help Center, accessed October 2026.
- Microsoft: About ads in Copilot, Microsoft Advertising Help, accessed October 2026.
- Google: ads in AI Overviews, Google Ads & Commerce Blog, May 21, 2025.
