On the one panel that publishes both channels on identical methodology, a LinkedIn click and a Google Ads click cost almost the same: $9.39 against $9.76. The leads do not. On the same panel, a LinkedIn lead cost $202 and a Google Ads lead $524.

That single pair of comparisons is more useful than most of what is written about this choice, and it is almost never the comparison presented.

This page covers what each platform publishes about its own costs, what the figures in circulation are actually worth, what the two channels are genuinely buying, and how to decide.

LinkedIn publishes no cost figure at all

Start here, because it explains why the numbers you have seen come from somewhere else.

The page that should answer the question. LinkedIn maintains a page titled “LinkedIn Advertising Costs & Pricing”, which asks “How much do LinkedIn ads cost?” and answers: “Your LinkedIn advertising costs are based on the type of activity you’re paying for and the ad auction.”

No number follows. Not an average, not a range, not a worked example. There is no official average cost per click, cost per thousand or cost per lead anywhere on the platform’s properties.

What it does document. The auction factors: you bid against advertisers targeting the same members, “the cost required to win the auction depends on the desirability of your target audience”, and an ad relevance score where “the more relevant your ad, the lower the price you pay”. Three bidding strategies exist: maximum delivery, cost cap and manual.

The minimum budgets, and where they actually live. $10 daily for any format and $100 lifetime for new campaigns, with a suggestion of $25 for new advertisers and $50 to $100 for existing ones. These appear on a marketing page, not in the help documentation, which says only that the minimum “is determined by the duration of your campaign”.

Why this matters more than it sounds. Every cost figure you have read about this platform was produced by a third party aggregating accounts. That is not disqualifying. It does mean the methodology question is the whole question.

The number everybody quotes, and where it stops

One figure dominates the English-language coverage of LinkedIn advertising costs. It does not survive being followed.

The figure. A cost per click of $5.26, usually presented alongside a CPM of $6.59 and a sponsored message cost of $0.80. The three travel together as a block, which is itself a clue: they are being copied, not measured.

Where the chain leads. Recent agency articles cite older agency articles, which cite a self-published post on a professional network dated June 2023. That post asserts the figures with no source, no sample size, no period and no country.

The detail that settles it. The publisher that appears to be the earliest reachable origin of the figure no longer uses it. Its current page gives a different range entirely, attributed to “hundreds of US-based marketing professionals”, with no sample count and no period.

A second tell. The same figure is dated 2021 in one repetition, undated in another, and 2024 in a third. A genuine measurement does not drift across three years while keeping two decimal places.

What to do with it. Do not quote it. If someone quotes it at you, ask for the sample and the period, and watch what happens.

Why this matters beyond one number. The same pattern produces most B2B advertising benchmarks: a plausible figure enters circulation, acquires citations, and becomes a fact through repetition rather than through measurement. The defence is always the same, and it takes two minutes: open the source, look for a sample size.

Citation chain behind the most widely quoted professional network cost per clickDiagram tracing the citation chain behind the cost per click figure most widely quoted for professional network advertising, and showing where that chain ends. The figure of five dollars twenty-six is usually presented alongside a cost per thousand impressions of six dollars fifty-nine and a sponsored message cost of eighty cents, the three travelling together as a block, which is itself an indication that they are being copied rather than measured. Recent agency articles cite older agency articles, which cite a self-published post on a professional network dated June two thousand and twenty-three, and that post asserts the figures with no source, no sample size, no period and no country. The detail that settles the matter is that the publisher which appears to be the earliest reachable origin of the figure no longer uses it, its current page giving a different range entirely, attributed to hundreds of United States based marketing professionals with no sample count and no period. A second indication is that the same figure is dated two thousand and twenty-one in one repetition, undated in another, and two thousand and twenty-four in a third, whereas a genuine measurement does not drift across three years while keeping two decimal places. The same pattern produces most business advertising benchmarks: a plausible figure enters circulation, acquires citations, and becomes a fact through repetition rather than through measurement.Following one number to its sourceRecent agency articles, dated 2024 and laterQuote $5.26 CPC, $6.59 CPM, $0.80 per message. Always the same three together.↓Older agency articlesSame block. One dates it 2021, another leaves it undated.↓A self-published post, June 2023No source. No sample. No period. No country. The chain ends here.And the publisher that appears to be its origin has since replaced it with a different range.The defence takes two minutes: open the source and look for a sample size.
A block of three figures travelling together, dated 2021 in one repetition and 2024 in another, ending at a post with no source. Source : MASTRATOS (2026)

The one benchmark with a declared sample

There is an exception, and it happens to allow the comparison this article is about.

What it publishes. Metadata releases its B2B advertising benchmark as a downloadable dataset under an open licence. The header states its scope plainly: 153 advertisers, $57.6M of 2025 ad spend, generated 19 August 2026, with a publication rule that no cut appears with fewer than five advertisers, under $50,000 of spend, or with any single advertiser above half the total.

The two lines that matter here. LinkedIn, on 138 advertisers: cost per lead $202, cost per click $9.39, CPM $63.19, click-through rate 0.67%, click-to-lead 5.9%. Google Ads, on 56 advertisers: cost per lead $524, cost per click $9.76, CPM $617.91, click-through rate 6.33%, click-to-lead 1.9%.

Why this comparison is worth more than a cross-source one. Same panel, same year, same definitions, same aggregation rules. Comparing a LinkedIn figure from one vendor to a Google figure from another compares two methodologies as much as two channels. Here that problem disappears.

The limit you must state alongside it. The panel is heavily weighted toward B2B technology, and its geography is not disclosed. It describes a population of advertisers, not the market, and certainly not you.

What it is genuinely good for. Order of magnitude, and above all the shape of the difference between the two channels, which is stable enough to reason about even if the levels are not yours.

Comparison of professional network and search advertising costs on a single declared panelTable comparing advertising costs on a professional network and on search, drawn from a single benchmark dataset covering one hundred and fifty-three advertisers and fifty-seven point six million dollars of two thousand and twenty-five advertising spend, published under an open licence with a rule that no cut appears with fewer than five advertisers, under fifty thousand dollars of spend, or with any single advertiser above half the total. The professional network line rests on one hundred and thirty-eight advertisers and shows a cost per lead of two hundred and two dollars, a cost per click of nine dollars thirty-nine, a cost per thousand impressions of sixty-three dollars nineteen, a click-through rate of zero point six seven percent and a click-to-lead rate of five point nine percent. The search line rests on fifty-six advertisers and shows a cost per lead of five hundred and twenty-four dollars, a cost per click of nine dollars seventy-six, a cost per thousand impressions of six hundred and seventeen dollars ninety-one, a click-through rate of six point three three percent and a click-to-lead rate of one point nine percent. The two channels therefore cost almost the same per click while differing by a factor of about two point six per lead, because the click-to-lead rate differs by a factor of about three. Because both lines come from the same panel, the same year, the same definitions and the same aggregation rules, the comparison avoids the usual problem of comparing two methodologies as much as two channels. The panel is heavily weighted toward business-to-business technology and its geography is not disclosed.One panel, two channels, 2025 spend153 advertisers, $57.6M of spend. Same definitions, same aggregation rules.METRICLINKEDIN (n=138)GOOGLE ADS (n=56)Cost per click$9.39$9.76Cost per lead$202$524Click-to-lead rate5.9%1.9%Click-through rate0.67%6.33%CPM$63.19$617.91Compare on clicks and the channels look equivalent. Compare on leads and they do not.The panel skews to B2B tech and does not disclose geography. Read the shape, not the levels.
The click comparison says the two channels are equivalent. The lead comparison says they are not. Both come from the same dataset. Source : Metadata B2B Ad Spend Benchmark (2026)

What each channel is actually selling you

The cost figures are a consequence. The cause is that the two platforms sell fundamentally different things, and the numbers above are exactly what that difference looks like.

Search sells an intent that already exists. Somebody typed the problem. You do not have to create the need, establish that the category exists, or interrupt anything. You compete on being the answer.

The professional network sells an identity. Somebody matches a firmographic profile, whether or not they are thinking about you today. You have to create the interest, which is why the click-through rate on the panel above is 0.67% against 6.33%.

Why the click-through gap does not mean what people say it means. A search click-through rate of 6.33% and a feed click-through rate of 0.67% are not comparable performance figures. One measures how well you answered a question, the other how well you interrupted a scroll. Almost every article comparing these platforms on click-through rate is comparing two unrelated quantities.

Why the click-to-lead gap runs the other way. Once someone clicks in a feed, having chosen to interrupt themselves, they arrive with more context about who you are, and the targeting has already filtered for whether they could plausibly buy. The panel shows 5.9% against 1.9%.

Where each one fails, and it is worth knowing before you pick. Search fails when nobody is searching, which on a narrow B2B category is common and unfixable by budget. The professional network fails when your proposition needs explaining to someone who did not ask, which is most propositions.

The framing that resolves most arguments. Search harvests. The professional network plants. A company with no harvest to bring in should not be planting first, and a company harvesting everything available and still short of pipeline has nothing left to harvest.

The comparison you will be shown, and why it misleads

Before the meeting where this gets decided, it helps to recognise the three framings that reliably produce the wrong answer.

“LinkedIn clicks cost five times more.” Usually built by comparing a LinkedIn figure from one source with a Google figure from another. On a single panel with identical definitions, the gap is $9.39 against $9.76. Most of the famous gap is a methodology difference wearing the costume of a market fact.

“LinkedIn’s click-through rate is terrible.” 0.67% against 6.33%, on that same panel, and the comparison is meaningless. One measures answering a question somebody asked, the other measures interrupting a scroll. No amount of creative work closes a gap that is structural.

“Google Ads has a better CPM.” The panel shows $63.19 against $617.91, which looks catastrophic for search until you notice that a search CPM is nearly worthless as a comparison: search impressions are scarce, targeted and answered on demand, and the whole point is that few of them go a long way.

The one comparison that carries information. Cost per lead, or better, cost per meeting held, on your own account, over a window matching your sales cycle. Everything else compares mechanisms rather than outcomes.

Why this happens so consistently. Each channel looks worst on the metric the other one is structurally good at. Anyone can produce a decisive-looking chart favouring either side by choosing which row to show. Ask which rows were left out.

What LinkedIn’s targeting is actually built on

The firmographic precision is the reason people pay these prices. It deserves a closer look, and LinkedIn documents it more honestly than most advertisers realise.

What is declared, not inferred. Job function and seniority are mapped from member-entered titles through a proprietary taxonomy, and LinkedIn states plainly: “no AI modeling or inference is used.” If you target by function and seniority, you are targeting declared data.

What is inferred. Company size “primarily uses Page admin account input data”, but “if a Page admin hasn’t entered a company size, then the company size is determined by the number of member accounts associated with the Page”. Industry can be extended: “additional industries may be inferred about the company”, via a deep learning model using company name, size, and members’ education and skills.

The edge case worth knowing. “If a member doesn’t enter a company name, LinkedIn may derive their current company based on their public IP address at ad serving time.”

On data freshness, there is no published metric. The only claim on the subject is marketing copy: members “have professional incentives to keep their profiles accurate and up to date”. That is plausible and it is not a measurement. Nobody publishes what share of profiles carry a current employer.

The operational constraints that shape campaigns. A minimum audience of 300 accounts to deliver at all, a cap of 200 companies in a company-name list, up to 300,000 through matched audiences, and reduced reach on several attributes in the European Economic Area and Switzerland.

How to use this rather than distrust it. Build targeting on the declared attributes where you can, function and seniority, and treat company size and inferred industry as approximations. The gap between the two is the gap between an audience you defined and an audience a model assembled.

Declared and inferred attributes in professional network advertising targetingTable distinguishing which targeting attributes on a professional advertising network rest on data members declared and which rest on inference, as documented by the platform itself. Job function and job seniority are mapped from member-entered job titles through a proprietary taxonomy, and the documentation states plainly that no artificial intelligence modelling or inference is used for either, so targeting by function and seniority targets declared data. Job title uses data that is primarily member input with some business logic and inference, and no third-party data. Company size primarily uses page administrator input data, but where an administrator has not entered a size, the company size is determined by the number of member accounts associated with the page, which is an estimate rather than a declaration. Company industry may be extended, additional industries being inferred about a company by a deep learning model that uses the company name, its size, and the education and skills of its members. An edge case documented by the platform is that where a member has not entered a company name, the platform may derive their current company from their public internet protocol address at the moment the advertisement is served. On the freshness of profile data the platform publishes no metric at all, the only statement on the subject being marketing copy asserting that members have professional incentives to keep their profiles accurate and up to date, which is plausible but is not a measurement, since nobody publishes what share of profiles carry a current employer.Declared, or assembled by a model?Job function and seniorityDeclaredMapped from member-entered titles. “No AI modeling or inference is used.”Job titleMostly declared”Primarily member account input, with some business logic and inference.”Company sizeFalls back to an estimateIf no admin entered it, size is derived from members attached to the page.Company industryCan be model-extended”Additional industries may be inferred”, by a model using name, size, skills.On profile freshness there is no published metric, only the claim that members have incentives to stay current.
Function and seniority are mapped from what members typed, with no inference. Company size falls back to a headcount estimate, and industry can be model-extended. Source : LinkedIn Help (2026)
What each advertising channel sells and how each one failsDiagram contrasting what each of the two main business advertising channels actually sells and the specific way each one fails. Search advertising sells an intent that already exists, since somebody typed the problem, meaning the advertiser does not have to create the need, establish that the category exists, or interrupt anything, and competes instead on being the answer. Professional network advertising sells an identity, since somebody matches a firmographic profile whether or not they are thinking about the advertiser today, meaning the advertiser has to create the interest, which is why the click-through rate on a common panel was zero point six seven percent against six point three three percent on search. The click-through gap does not mean what is usually claimed, because a search click-through rate measures how well a question was answered while a feed click-through rate measures how well a scroll was interrupted, so comparing the two compares unrelated quantities. The click-to-lead gap runs the other way, at five point nine percent against one point nine percent, because someone who chooses to interrupt themselves arrives with more context and the targeting has already filtered for whether they could plausibly buy. Search fails when nobody is searching, which is common on a narrow business category and cannot be fixed with budget. The professional network fails when a proposition needs explaining to somebody who did not ask, which is most propositions. Search harvests and the network plants.Two channels, two failure modesSearch sells intentSomebody typed the problem.You compete on being the answer.It harvests.It fails when nobody is searching,which no budget can fix.The network sells identitySomebody matches a profile.You have to create the interest.It plants.It fails when the proposition needsexplaining to someone who did not ask.Establish whether you have a harvest before deciding you need to plant.And do not split a small budget across both to test. Two underfunded campaigns teach you nothing.
Establish whether you have a harvest before deciding you need to plant. Most companies get this order wrong. Source : MASTRATOS (2026)
Decision sequence for choosing between search and professional network advertisingDiagram setting out the decision sequence for choosing between search advertising and professional network advertising in business markets, with the test that answers each step. The first step is to write down the close rate from lead to customer and the average deal value, because no channel comparison means anything without them, a channel being cheap only relative to what a customer is worth. The second step is to establish whether a harvest exists, by running search alone for six to eight weeks on exact and phrase match with a tight negative keyword list, the question being whether measurable intent exists in the category rather than whether the campaign is optimised. The third step is to read the search terms report rather than the dashboard: if the arriving queries are people describing the problem in their own words there is demand, whereas if they are students, job seekers and adjacent categories there is keyword volume without demand, which is a different and more useful finding. The fourth step is to consider the professional network only then, with a funded budget on a genuinely narrow audience, one offer, and a clear reason for somebody who was not looking to stop and read. The metric to measure at the end is the cost per meeting held rather than the cost per lead, since it is the earliest metric reflecting lead quality and will differ between the two channels far more than cost per lead does. The two channels must be judged on different clocks, because a quarter tells whether search has a harvest but planting shows up on the timescale of the sales cycle.A sequence, not a ranking1. Write down your close rate and deal valueIf you cannot, stop here. No comparison means anything without them.2. Run search alone for six to eight weeksExact and phrase, tight negatives. One question: does measurable intent exist?3. Read the search terms report, not the dashboardPeople describing your problem means demand. Students and adjacent categories means volume.4. Only then, fund the network properlyNarrow audience, one offer, a reason for someone who was not looking to stop.Measure cost per meeting held, not cost per lead. And judge the two on different clocks.
Not a ranking. A sequence, and every step has a test you can run rather than an opinion you can hold. Source : MASTRATOS (2026)

How to actually decide

The choice is not a ranking. It is a sequence, and the sequence depends on facts about your market you can establish in an afternoon.

Start with the question that settles most cases. Is anybody searching for what you sell? Open the search terms report of any existing account, or run a small search campaign for a month. If there is intent, harvest it before you do anything else, because it is the cheapest demand you will ever meet.

If there is little or no search volume. That is common on narrow B2B categories, and it is not a failure of your keyword research. It means the demand has to be created before it can be harvested, which is what the professional network is for.

If search volume exists and you are already taking all of it. This is the second, better reason to add the network: you have exhausted the harvest and need to plant. It is also the point at which most companies conclude, wrongly, that “search stopped working”.

The deal value test. A cost per lead in the low hundreds only makes sense against a deal worth thousands, with a close rate you can actually state. If you cannot name your close rate from lead to customer, you cannot judge either channel, and that is the work to do first.

The narrowness test. Firmographic targeting earns its premium when your target genuinely is a list of a few hundred companies. If your target is “any business”, you are paying for precision you are not using.

What not to do. Split a small budget across both to “test”. Two underfunded campaigns produce two inconclusive results and a quarter lost. Fund one properly, learn something, then decide.

Settling it on your own account in one quarter

Benchmarks orient. They do not decide. Here is the smallest experiment that actually decides, and it fits in a quarter.

Step one, before spending anything. Write down your close rate from lead to customer, and your average deal value. If you cannot, stop here and fix that: no channel comparison means anything without them, because a channel is only cheap relative to what a customer is worth.

Step two, establish whether a harvest exists. Run search alone for six to eight weeks, on exact and phrase match, with a tight negative list. You are not optimising yet. You are answering one question: does measurable intent exist in this category?

Step three, read the search terms report rather than the dashboard. If the queries arriving are people describing your problem in their own words, you have demand. If they are students, job seekers and adjacent categories, you have keyword volume without demand, which is a different and more useful finding.

Step four, only then consider the network. With a funded budget on a genuinely narrow audience, one offer, and a clear reason for somebody who was not looking to stop and read.

What to measure at the end, and it is not cost per lead. Cost per meeting held. It is the earliest metric that reflects lead quality, it uses a denominator you control, and it is the one that will differ between the two channels far more than cost per lead does.

The honest caveat about timing. A quarter tells you whether search has a harvest. It does not tell you whether the network works, because planting shows up on the timescale of your sales cycle. Judge them on different clocks or you will cut the second one before it reports.

Where to go next

You want to structure the search side. Google Ads account structure for B2B.

You are splitting a budget between platforms. How to split budget between Google and Meta in B2B.

Your cost per lead is the number in dispute. Cost per lead.

You want the cost of a customer rather than a contact. How to calculate CAC, and CAC at low volume.

You are setting the budget in the first place. Small business advertising budget.

You want the buying mechanics behind both. Media buying explained.

In short

  • LinkedIn publishes no cost figure. Its own pricing page answers “how much do LinkedIn ads cost?” with a sentence about auctions and no number.
  • The $5.26 CPC everyone quotes is untraceable, and the publisher that appears to be its origin has replaced it. Do not use it.
  • One benchmark declares its sample: 153 advertisers, $57.6M of 2025 spend, published under an open licence.
  • On that panel, clicks cost almost the same: $9.39 on LinkedIn against $9.76 on Google Ads.
  • Leads do not: $202 against $524, because click-to-lead ran 5.9% against 1.9%.
  • The click-through comparison is meaningless. 0.67% against 6.33% compares interrupting a scroll with answering a question.
  • Function and seniority are declared data, with no inference. Company size falls back to a headcount estimate, and industry can be model-extended.
  • Search harvests, the network plants. Establish whether you have a harvest before deciding you need to plant.

Two channels that fail in different ways are not ranked against each other, they are sequenced. Book a diagnostic, or see how we approach B2B paid acquisition.