We enriched 1,700 demand generation agencies against a single question buyers are told to ask and rarely can. Do they run demand generation for themselves? For roughly 7 in 10 of them, the answer is no. The industry that sells demand as infrastructure mostly operates as undifferentiated production, at the exact moment the market stops paying for undifferentiated production.
| Signal measured across 1,700 agencies | Share of the field |
|---|---|
| Run no real demand engine for themselves (none or thin, ads-reliant) | 70% |
| Get under 1,000 organic visits a month to their own site | 71% |
| Publish no original research or proprietary data | 87% |
| Position as generalist or shallow on verticals | 69% |
| Carry no published pricing | 75% |
| Carry three or more red flags at once | 66% |
| Clear the operator bar on every criterion at once | 2.8% |
This is a first-hand data report. It reads one population against one fixed set of criteria, and it asks what that data says about the discipline, the model behind it, and where the model is heading.
The population is 1,700 demand generation agencies, deduplicated by domain, sourced from Clutch, Crunchbase, and Sales Navigator, then enriched against a fixed schema. Every agency in the set resolved to a live website and cleared a basic identity check, so lead-list vendors, appointment-setters, dead domains, and non-agencies were removed before this analysis. What remains is the working field a B2B buyer would actually encounter when shopping for a demand gen partner.
Each agency was scored on the criteria a B2B buyer is advised to use when choosing a demand generation partner: real specialization, demand proof, process transparency, named senior talent, and honest economics. We added the layer few buyers can run at scale, an assessment of whether the agency practices the discipline it sells: does it rank organically for what it does, does it publish original research, does it have genuine vertical depth or just a couple of named logos.
This report uses two units on purpose. Percentages mark concrete counts inside the dataset: the share of agencies with a given tag, a red flag, a score band. Ratios mark the reader-facing gist, where a proportion is easier to hold as "roughly 7 in 10" than as a decimal.
Some figures rest on a thinner slice and are flagged Directional inline. The organic-traffic bands come from third-party estimates and undercount small sites. Pricing and engagement data are sparse, because most of the field publishes neither. The critical-assessment layer is machine-judged from public signals, so it reads a site as a buyer would on a first pass, not as an auditor would with access to the books. Named leaders are ranked on our composite score and should be read as a well-evidenced shortlist, not a settled league table.
The headline is not that demand generation agencies are bad at marketing. It is that the field is built around a job that is losing its value, and most of the field has not noticed.
For two decades the agency sold execution. It produced the content, ran the ads, built the pages, and sent the reports. That work was scarce enough to be worth paying for. AI removed the scarcity. When a machine drafts the copy, builds the variants, and optimizes the media, the production layer stops commanding a premium, and the buyer starts asking what is left. Our data says that for most agencies, not much is left that a client cannot now assemble in-house.
Six findings carry the report.
46% show no real demand engine and another 24% run something thin or ads-reliant. Only about 1 in 10 run a strong own inbound motion.
Jump to evidenceUnder 1,000 organic visits a month to their own domain, and only 3% clear 1,000. The people selling a way off the paid-and-referral treadmill are on it themselves.
Jump to evidenceThe asset that earns citations and AI answers is published by almost no one; only 2% publish anything substantial. It is the capability the next cycle rewards most.
Jump to evidenceOnly 8% show genuine vertical depth, and 14% make vertical claims their content does not support. The typical agency lists nearly five service lines.
Jump to evidenceAmong agencies with a public Clutch score, 75% sit at 4.9 to 5.0 and the median is a perfect 5.0. Underneath it, 44% show few or no case studies.
Jump to evidenceThe highest scorers run their own demand engine, own a vertical, and publish research. The separation tracks operating model, not headcount.
Jump to evidenceEverything that follows works from that constant. Each section states a pattern, shows where the 1,700 fall on it, and reads what it says about the agency model in 2026.
Before quality comes structure, and the structure of this market is a long tail of very small firms.
61% of these agencies employ 10 people or fewer, and 87% employ 50 or fewer. Only 132 firms in the entire set, under 1 in 12, are larger than 50 people. This is a cottage industry, not a consolidated one, and the fragmentation matters because it shapes what most agencies can actually deliver. A team of six selling "full-service demand generation across paid, content, SEO, ABM, and web" is describing an ambition, not a capability.
The field is also young and turning over. 32% were founded in 2023 or later, and 51% since 2020. Only 16% predate 2010. A buyer shopping this market is mostly meeting firms with a short track record, in a category where the proof that matters, a multi-year demand system that compounded, takes years to accumulate.
Geographically it concentrates where B2B software money is. Of the agencies with a known headquarters, 37% sit in the United States, 12% in the United Kingdom, and 7.5% in India, with Canada and Australia next. The offshore share is meaningful and clusters at the low-cost, execution-for-hire end of the field.
The one structural fact that frames the rest of the report is self-positioning. Where a primary category is discernible, "full-service" is the single largest bucket at 47%, more than twice the next category. Only 16% lead with demand generation as their actual specialty. Most of this market is generalist shops wearing a demand gen label because that is what the buyer is searching for.
Fragmentation is the buyer's first filter, not a footnote. A sub-10-person shop claiming five service lines is telling you where the work will actually go, to freelancers and junior generalists, not to the named senior operator you are hiring for. Size is not the disqualifier. The mismatch between size and claimed scope is.
A firm that does everything specializes in nothing. It is the first thing a careful buyer checks, and the field fails it at scale.
Among agencies that list their service lines, the typical firm names nearly five of them, and 72% name four or more. The lines sprawl across demand gen, paid media, SEO, content, social, email, web development, ABM, CRO, and branding, in almost every combination. This is the "integrated storytelling" positioning the guide flags, a wide surface area that reads as capability and functions as hedging.
The problem is not breadth for its own sake. It is that demand generation is a specific operating discipline, building content and campaigns as decision infrastructure that moves a buying committee from unaware to qualified, wired to the CRM and measured in pipeline. That is not the same job as running Google Ads or shipping blog posts, and an agency that treats it as one line item among ten is not set up to run it.
The data underneath the positioning confirms the gap. When we tag what each agency can actually evidence rather than what it claims, the demand gen specifics, ABM, lifecycle and lead routing, decision-stage content, attribution, thin out fast. The claim is near-universal. The proof is not.
This is the spine of the report, and the cleanest signal in the dataset. The most reliable evidence that an agency can build a demand system is that it runs one for itself, in public, where anyone can check.
Most cannot. Judged on their own marketing integrity, 46% show no real demand engine at all and another 24% run something thin or ads-reliant. Together, 70% of demand generation agencies do not practice demand generation on their own behalf. Only about 10% run a strong own inbound motion, the thing they are selling.
The traffic data says the same from a second angle. 71% of these agencies draw under 1,000 organic visits a month to their own site, and only 3% clear 1,000 (Directional, third-party estimates undercount small sites, which only sharpens the point). An industry whose entire premise is manufacturing inbound demand generates almost none of it for itself.
| Own demand engine | Share of field | Break 1,000 organic visits |
|---|---|---|
| Strong own inbound | 10% | 24% |
| Moderate | 20% | 3% |
| Thin or ads-reliant | 24% | 0% |
| None or absent | 46% | 0% |
Read the right-hand column down. Even among the agencies we rated strongest on their own marketing, only a quarter break a thousand organic visits. Below that top tier, organic demand effectively does not exist. The field is running on referrals, outbound, marketplace listings, and paid, the exact dependencies its clients hire it to escape.
There is an honest counterargument, and it deserves stating. An agency can be good at a job it does not perform for itself, the way a chef need not eat at home. But demand generation is not a craft performed once and handed over. It is a compounding system, and the only way to prove you can build one that compounds is to point at one that did. The 70% cannot. The 10% can, and as the next sections show, they are the ones pulling away.
Before the case studies, before the pitch, look at the agency's own site the way a buyer looks at a prospect. Does it rank for anything a buyer would search. Does it publish, consistently, for a defined audience. Does it capture and nurture, or just collect a "contact us." An agency that has not built its own demand engine is selling you a blueprint it has never built from.
The same criteria we ran across 1,700 agencies apply to any B2B company. A Content RevOps audit scores your presence, conversion architecture, proof layer and answer-engine readiness in a week.
Original research is the asset that earns links, citations, and inclusion in AI answers. It is also the one the field almost never produces.
87% of these agencies publish no original research or proprietary data of any kind. 11% publish some, and 2% publish anything substantial. In a market where every competitor can now generate fluent content on demand, proprietary data is the clearest remaining line between a firm with a point of view and a firm rephrasing everyone else's.
This scarcity is not a craft failure. It is a business-model failure. Original research costs time and judgment before it returns anything, which a billable-hours shop optimizing utilization will never prioritize. So the capability that most durably separates an agency, and most reliably earns the citations that AI assistants now surface, is the one the dominant economic model actively discourages.
The consequence shows up downstream. An agency with no proprietary data has nothing distinctive to rank for, nothing for the answer engines to quote, and nothing to anchor genuine thought leadership. It is left competing on execution and price, in the one year execution and price stopped being defensible.
Specialization is the differentiator buyers are told to weigh first, the wider market now agrees is the surviving one, and the field mostly fakes.
Only 8% of these agencies show genuine vertical depth, a real hub of vertical content, multiple clients in the space, and research or proof specific to it. 49% read as outright generalist and another 20% as shallow, a couple of named logos standing in for expertise. 14% make vertical claims their own content does not support, the "we serve SaaS, fintech, healthcare, and manufacturing" banner with nothing underneath it.
Vertical depth is not a marketing preference in B2B. It is what lets an agency speak a buying committee's language, anticipate its objections, and produce content a specialist would trust with a complex decision. A generalist can write around a category. Only a specialist can write from inside it, and buyers in trust-led sales can tell the difference in the first paragraph.
Where depth does exist, it clusters. B2B SaaS is the deepest-served vertical by a wide margin, with manufacturing, professional services, and fintech behind it. The regulated and technical verticals where content is hardest to fake and most valuable, life sciences, medical device, legal, construction, are thinly served. The harder the vertical, the fewer agencies can actually operate in it, which is precisely where the opening is.
Test the vertical claim in one move. Ask for the vertical's content hub, not the logo wall. A real specialist has a body of work a buyer in that industry would bookmark. A generalist has a case study and a claim. The gap between the two is the difference between an agency that will shorten your sales cycle and one that will lengthen your review.
Buyers are told to verify agencies through proof and third-party validation. Both signals have degraded to the point of near-uselessness, in opposite directions.
Start with reviews. Among the agencies carrying a public Clutch score, 75% sit between 4.9 and 5.0, 93% sit at 4.7 or above, and the median score is a perfect 5.0. A rating that nearly everyone holds at the ceiling carries no information. It cannot rank, it cannot warn, and a buyer using it as a filter is filtering on noise. Meanwhile 63% of the field surfaces no external rating at all, so the choice a buyer actually faces is between an uninformative perfect score and no score.
The owned-proof layer is thin in the other direction. 44% of the field shows few or no case studies, and detailed, revenue-backed case studies, the kind that report SQLs, pipeline, CAC, or cycle impact rather than "increased traffic," are rarer still. The proof buyers are told to look for exists on a minority of sites, and the proof that reports commercial outcomes rather than activity is rarer than that.
Transparency compounds the problem.
Put together, the two layers a buyer is told to lean on both fail. The third-party score is saturated and uninformative. The owned proof is missing or activity-based. The trust market for demand gen agencies is running on badges, not evidence, which is the same zero-trust condition their clients' buyers are in, one level up.
Read individually, these warning signs are common. Read together, they are the norm.
These are not isolated. The average agency in this dataset carries just over three of these flags at once, the median is three, and 66% carry three or more. Only 27 agencies in the entire set of 1,700, about 1.6%, carry none.
That is the field in one number. A buyer choosing at random is overwhelmingly likely to meet a firm that hides its pricing, has no content footprint of its own, and cannot point to real case studies or a named senior team. The flags cluster because they share a root cause. A small, generalist, execution-for-hire shop optimizing billable hours has no structural reason to build the things the flags measure, so it builds none of them.
The flags are not a scoring exercise, they are a coherence check. Pricing, content footprint, case studies, named seniors, and vertical depth are the visible outputs of an agency that runs itself as a system. Their absence, all at once, is not five separate gaps. It is one, the absence of the system.
For all the weakness in the field, a clear top tier exists, and what defines it is precise. The agencies that score highest are the ones that do for themselves exactly what they sell.
The separation is stark on every axis we measured. Scored out of 100 on the full criteria set, the field averages 48 and the median is 46. But the score splits cleanly by operating model.
Running your own demand engine is worth 44 points of score over running none. Owning a vertical is worth 34. Publishing real research is worth 35. These are not independent tactics. They are three faces of one capability, an agency operated as a compounding system rather than a billable-hours shop, and they rise and fall together.
The top decile makes the pattern unmistakable. Of the 170 highest-scoring agencies, 99% run their own marketing, 79% have real vertical depth, and 61% publish original research, against field rates of 30%, 31%, and 13%. The leaders are not doing one thing well. They are running the whole operating model the rest of the field skips.
The highest-scoring agencies in the dataset are a recognizable set, and their profiles rhyme. Every one runs a visible demand engine of its own, most own a vertical, and the majority publish proprietary research. The table reads them at a glance (Directional, ranked on a composite of public signals, not a settled league table).
| Agency | Score | Own engine | Vertical | What earns the score |
|---|---|---|---|---|
| INFUSE | 92 | Strong | Moderate | Global demand-gen firm running its own research and inbound at scale |
| Hinge | 92 | Strong | Deep | Annual research study for professional-services marketing, ranks for the terms it sells |
| RevPartners | 91 | Strong | Moderate | RevOps and GTM engineering on HubSpot, strong own inbound |
| Refine Labs | 91 | Strong | Deep | Demand-led B2B SaaS specialist with a widely cited point of view |
| Walker Sands | 90 | Strong | Deep | Integrated brand-to-demand for B2B tech, long-running original research |
| Discovered Labs | 90 | Strong | Deep | SEO and answer-engine optimisation run as one engine, reported on pipeline |
| Ironpaper | 88 | Strong | Moderate | B2B-only, built for long complex cycles, publishes its own market research |
| AGENCY (UK) | 88 | Strong | Deep | Healthcare-only demand generation focused on the middle of the funnel |
| Overthink Group | 88 | Strong | Moderate | Publishes quarterly SERP and AI-citation research, ranks for its category terms |
| Miha Cacic Marketing | 88 | Strong | Deep | Comparative content that outranks the SaaS brands it writes about |
| Sagefrog | 86 | Strong | Deep | Life-sciences, medtech, and manufacturing depth with its own inbound |
| Directive | 86 | Strong | Moderate | B2B SaaS performance with strong own search visibility and thought leadership |
Look past the names and the leaders cluster into three operating patterns, none of which the median agency runs.
The research houses. Firms like Hinge, Ironpaper, Walker Sands, Overthink Group, and memoryBlue turn proprietary data into their own demand. They publish an annual or quarterly study, rank for the category it seeds, and earn the citations that feed both search and AI answers. The research is not their content marketing. It is their demand engine, and it is the single capability 87% of the field does not have.
The vertical specialists. AGENCY in healthcare, Percepture and Sagefrog in life sciences, New Perspective in manufacturing and cleantech, and Obility in B2B SaaS go deep enough in one industry that a buyer inside it trusts them on sight. Their content reads as written from inside the category, not around it, which is exactly the depth the 92% cannot show. The harder and more regulated the vertical, the fewer competitors reach it, so depth doubles as a moat.
The dogfooders. Miha Cacic Marketing, Overthink Group, Discovered Labs, and Impactable win their own category's search and use that as the proof. When an agency outranks the SaaS brands it writes comparison content about, or ranks first for the answer-engine terms it sells, the case study is the search result itself. It is the cleanest sales asset in the field, and almost no one can produce it.
The bar is high because so few clear it. Only 2.8% of the field, 47 agencies out of 1,700, meet the full standard at once: vertical depth, a real own-marketing motion, original research, a named senior team, and results-based proof. And the firms that clear it are not the biggest in the set. Several run teams of 1 to 10 people, and several of the largest agencies in the dataset clear none of it. The separation is an operating choice, not a headcount.
The good news for a buyer is that the shortlist is short and knowable. The lesson for an agency is that the operator model is not crowded. It is nearly empty.
Step back from the 1,700 and the shape of the whole model comes into view, along with the reason it is now under pressure.
The field this data describes is a direct product of the old economics. When execution was scarce, a small generalist shop could sell hours across many services, keep its pricing opaque, staff delivery with juniors, and never need a demand engine, a research capability, or a vertical of its own, because the work itself was the scarce thing. Every dominant pattern in this report, the sprawl, the missing own-marketing, the absent research, the shadow teams, the opaque pricing, is what that model rationally produces.
That model is now being repriced in real time, and the external signals are consistent. Agency profit margins have fallen from around 30% in the industry's better years to roughly 10% today (VoxComm and Lodestar, 2026). In 2025, global ad spend rose 8.6% while the big holding companies' revenue fell 1.2%, as platform automation captured the execution work agencies used to intermediate (eMarketer, 2026). Forrester's 2026 survey finds 93% of B2B marketers still use agencies, but the share expecting to increase agency spend fell 13 points year over year, with the expected increase for content-creation work dropping from 41% to 26%. Buyers are not firing their agencies. They are quietly narrowing what they pay them for.
The mechanism is AI, and it hits the field exactly where the field is weakest. AI commoditizes execution by making drafting, variant production, media optimization, and reporting abundant and near-free. What it cannot commoditize is proprietary evidence, genuine vertical judgment, credible third-party authority, and a demand system that compounds, the precise capabilities that, on our data, 7 to 9 out of 10 agencies lack. As one industry framing puts it, production commoditizes, judgment does not, and clients already rank "the work" far below strategic and business understanding as the reason they value a partner (VerityRI, 2025).
The in-housing data closes the loop. More than 80% of large brands now run an in-house agency of some kind (ANA), the hybrid model is now the most common operating setup for B2B marketing at 46% and rising (Sagefrog, 2026), and 15% of B2B marketers cut agency spend outright in 2025 because AI let them bring the work in-house (Marketing Week, 2025). When the execution an agency sold is the execution a client can now run internally with AI, the agency has to be worth more than the execution. Most of this field is not yet built to be.
The gap between the field and its future is not a list of tactics. It is a set of capabilities the old model never needed and the new one requires.
The agencies that build these will not be selling execution with AI bolted on. They will be selling the operating model the execution used to hide inside, and charging for judgment, evidence, and outcomes rather than hours. The rest will compete on price against software, which is not a competition.
The dataset is 1,700 demand generation agencies, deduplicated by domain from Clutch, Crunchbase, and Sales Navigator. Every agency resolved to a live website and passed a for-profit B2B agency identity check, which removed lead-list vendors, appointment-setters, dead or parked domains, and non-agencies before analysis.
Each agency was enriched against a fixed schema of 55 fields, then scored on nine critical-assessment criteria covering specialization, own-marketing integrity, original research, proof, transparency, named senior talent, and red flags.
The critical layer, own-marketing integrity, vertical depth, original research, transparency, and red flags, is judged from public signals, the way an informed buyer would read a site on a careful first pass. It measures visible operating maturity, not guaranteed delivery quality, so a strong agency with a weak website scores below its true ability.
Organic-traffic bands come from third-party estimates that undercount small sites and should be read as magnitude, not precision. Pricing data is sparse because most of the field publishes none. The named leaders are ranked on public signals and offered as a starting point, not a final ranking. All figures are a point-in-time snapshot, collected in mid-2026.
The pattern across 1,700 agencies is the same one we find across the companies that hire them. The execution layer is built, and the operating layer that turns it into compounding pipeline is not. Content RevOps builds content as revenue infrastructure, wired to CRM, automation, and reporting, anchored in original research and vertical depth, and measured in pipeline rather than activity. Our case studies are named and deep-linked, from King's College London to Ori Learning to Heliogen, because in a market this saturated with badges, evidence is the only thing left that a buyer trusts.
The Demand Generation Agency Landscape 2026. A first-hand read of 1,700 agencies against the criteria that separate operators from storytellers. August 2026. Percentages mark counts in the dataset; ratios mark the reader-facing gist; directional findings flagged inline. © 2026 Content RevOps.