Clay marketing teardown: a measured audit of its content, ads and video
Everyone retells Clay's marketing from interviews and conference talks. We captured what the company actually published, and read it.
21 min readWritten by Ampere
Contents
- What we found
- Most of clay.com is not the blog
- What seventeen thousand pages say
- LinkedIn runs in two registers
- Google and LinkedIn are selling different companies
- The channel that looks dead and isn't
- The last six months, counted
- Publish the output, teach the work
- What actually transfers
- How we did this research
Clay sells the plumbing underneath business-to-business sales. You give it a list of companies, it pulls in facts about them from a few hundred data providers, runs AI agents over the result, and pushes what it finds into the tools a sales team already uses. In early 2025 the company described itself as "a data enrichment tool, a way for GTM teams to get cleaner lists" — go-to-market, the umbrella term for sales, marketing and the operations between them. It now says it helps people "build systems that grow revenue across data, agents, orchestration, execution, and governance."
The escalation tracks the company. Clay was founded in 2017, spent six years on product, and then, by its own account, went from $1M to $100M in annual recurring revenue in two years. It raised $100M at a $3.1B valuation in August 2025, and a January 2026 employee tender priced it at $5B. Along the way it did something rarer than growing fast: it named a job. The GTM engineer, the person who automates the manual work nobody on a revenue team wants to touch, went from Clay's coinage to a title on real job postings.
All of which makes Clay worth taking apart. It is the reference example founders point to when they argue a category can be created rather than entered, and its marketing gets retold constantly: as a community story, a founder-led-content story, a category-design story. Those retellings share a weakness. They are built from what Clay says about its marketing, in interviews and conference talks, and almost never from what Clay's marketing is.
So we went and got it: every public marketing artifact we could reach across nine of Clay's surfaces. 17,323 URLs on its own domain, 67 LinkedIn posts with engagement attached, 440 LinkedIn ad creatives, 118 Google ad creatives carrying 287 text variations, 163 YouTube videos, 249,086 words of video transcript, its entire Instagram history, and 125,219 archive records establishing when pages first appeared. Then we read it.
What we found
The short version, before the evidence.
- Clay's blog is 1.4% of its site. Three programmatic directories (templated pages generated from data, not written) are 87.6%.
- The largest of those directories, at 10,345 pages, answers four questions: email format, funding, executives, offices.
- A second directory, 3,335 pages, is one personalized landing page per target company.
- On LinkedIn, 52% of ad creatives invite you to a live workshop or stream. On Google, 1% does.
- YouTube holds 249,086 words of transcript, and its dominant format is a numbered course with named frameworks.
- That channel has published two videos since October 2025, and they are its two longest broadcasts ever.
Most of clay.com is not the blog
Start with the thing Clay owns outright. Its sitemap lists 17,323 URLs, captured August 3, 2026. The blog is 238 of them, 1.4% of the domain.
The bulk sits in three programmatic directories — pages generated from a template and a data source rather than written one at a time. /dossier holds 10,345 pages, /gift holds 3,335, and /integrations holds 1,493. Together they are 87.6% of every URL on the domain.
| Total sitemap URLs | 17,323 |
|---|---|
| /dossier | 10,345 (59.7%) |
| /gift | 3,335 (19.3%) |
| /integrations | 1,493 (8.6%) |
| /blog | 238 (1.4%) |
| Everything else | 1,912 (11.0%) |
The archive record shows these were not assembled gradually. /dossier went from 37 archived pages in July 2024 to 7,113 by May 2025, the bulk of it inside about ten months, then essentially flat for a year.
pages first archived
View the data
| Period | pages first archived |
|---|---|
| 2024-07 | 37 |
| 2024-08 | 217 |
| 2024-09 | 546 |
| 2024-10 | 1011 |
| 2024-11 | 1219 |
| 2024-12 | 1812 |
| 2025-01 | 2913 |
| 2025-02 | 3167 |
| 2025-03 | 3890 |
| 2025-04 | 5898 |
| 2025-05 | 7113 |
| 2025-06 | 7924 |
| 2025-07 | 7934 |
| 2025-08 | 7944 |
| 2025-09 | 7949 |
| 2025-10 | 7955 |
| 2025-11 | 7959 |
| 2025-12 | 7977 |
| 2026-01 | 7985 |
| 2026-02 | 7987 |
| 2026-03 | 7991 |
| 2026-04 | 7991 |
| 2026-05 | 7992 |
| 2026-06 | 7998 |
| 2026-07 | 8022 |
| 2026-08 | 8023 |
Source: Internet Archive CDX API, first-capture timestamps joined against clay.com's sitemap. Captured 2026-08-03. First capture is a lower bound on publication date, so the true curve sits at or to the left of this one. About a fifth of today's dossier pages have no archive capture at all, which is why the curve tops out below the current total.
/gift is newer and steeper. Its first archived page appears in November 2025; by May 2026 there are 2,878. Nothing to three thousand pages in seven months.
The archive also holds 4,848 pages under /number-of-employees that no longer appear in Clay's current sitemap, plus 608 under /customer-pages and 361 under /experts. Clay has built programmatic directories, and Clay has retired them. What survives, presumably, is the set that earned its keep; the archive shows the retirements, not the reasons.
What seventeen thousand pages say
Counting pages is the easy half. The directories are templated, so a sample tells you what all of them contain, and the contents are more interesting than the totals.
Take /dossier, the largest at 10,345 pages. Every URL ends in a question type, so the whole population can be classified from the slugs rather than a sample.
| Email format (…-email-format) | 5,028 pages (48.6%) |
|---|---|
| Funding and investors (…-funding) | 2,534 pages (24.5%) |
| Executives and leadership (…-ceo, -cto, -cfo, -executives) | 2,235 pages (21.6%) |
| Office locations (…-locations) | 544 pages (5.3%) |
| Everything else | 4 pages |
Four question types cover 10,341 of 10,345 pages. The titles read like the questions themselves: Who is the CEO of 0x in 2026? Will Warren's Bio. How Much Did Altruist Raise? Funding & Key Investors. Australia Post Email Format 2026 — Find the Right Pattern.
Now look at what Clay sells. Email addresses, funding events, executive names and company locations are what its enrichment product returns when you point it at a company. The directory is the product's output, published one row at a time, closing with a call to action that reads "Get access to more people and company data using Clay."
The largest content asset Clay owns is a public sample of what its product does.
The machine is thorough enough to have indexed its own maker. clay.com/dossier/clay-funding is a real page, in the sitemap, built from the same template as the other 10,344 — How Much Did Clay Raise? Funding & Key Investors — with the same "Use Clay to Get Funding Data" block at the bottom. On our capture date, searching for Clay's funding history returned it.
/gift is the stranger directory, and the more instructive. Every one of the twenty pages we sampled carries the same headline: Michelangelo had marble, Marketing has Clay. Each slug is a company name — 10x-genomics, 1worldsync, 360learning. Each page names customers that particular company would find familiar: one greets a life-sciences target with "close more customers like Broad Institute, Memorial Sloan Kettering Cancer Center, and Allen Institute," while an apparel target gets "VF Corporation, The North Face, and New Balance."
This is a personalized landing page generated per target account, at a scale of three thousand. And Clay publishes the recipe. On the same domain sits a public template titled Auto-generate personalized landing pages for all prospects.
The rest of the surface follows the same logic. /integrations is 1,493 pages, one per action Clay can take in another tool: Enrich a company's investors with Crunchbase, Add contact to sequence with Apollo.io. /tools is 237 pages of How to find company growth rate with AI. /glossary is 268 definitions, some well outside Clay's category, including What is Docker? and What is Cross-Site Scripting?
Then there is the blog everyone writes about. Of 70 sampled posts, 45 carry a year in the title: AI Lead Scoring: Definition, Benefits & How It Works 2026, B2B Cold Email Deliverability: 21 Best Practices 2026. That is the signature of an evergreen library maintained on a refresh cycle, which is what the April 2026 date anomaly showed from the other direction. Roughly a third are how-to guides and a sixth are comparison and alternatives pages.
Off its own domain, the register changes completely.
LinkedIn runs in two registers
Clay's LinkedIn page has 179,890 followers. In the 41 days we could reach, June 24 to August 3, 2026, the company published 67 posts, a rate of 11.4 per week, hitting six posts in a single day at peak. Those 67 posts drew 15,028 reactions and 2,114 comments. Not one of them has zero comments.
The median post drew 125 reactions; the mean is 224. That gap is the signature of a channel carried by outliers, and the outliers are not what a content strategy deck would predict.
| Brand / humor (26 posts) | 341 average reactions |
|---|---|
| Product announcements (9 posts) | 223 average reactions |
| Events and webinars (11 posts) | 149 average reactions |
| Hiring (4 posts) | 135 average reactions |
| Education and tutorials (17 posts) | 115 average reactions |
One caveat we return to in the ads section: some of these posts also ran as ads, and public data cannot say which reactions were bought.
The best-performing post in the window is the words "you ever just" followed by a scream of roughly two hundred letter As. It drew 1,372 reactions and 143 comments. Second is a fake sports-trade bulletin announcing that LeBron James "will not return for Q3 and will instead take his talents to the Philadelphia 76ers," at 1,354. Third, at 1,153, is the same joke run two weeks earlier, with LeBron leaving the Lakers to become "a Forward Deployed GTME."
Read the posts as text and a second pattern appears that the engagement figures hide. Post length is bimodal — clustered at two extremes with little in between. Of 65 posts with text, 11 run under 120 characters and 33 run over 600, and the median is 618.
The short mode is pure play: "pov: you locked tf in." "2020 was 2020 days ago." "clocking in for my 9-5 before my 5-9." The long mode is dense product writing, formatted with numbered lists and bold Unicode headers, the kind that names eight releases and explains each. Very little sits between them: no medium-length think-pieces, no reformatted blog introductions.
The other tell is what is missing. Only 7 of 65 posts carry a link. Clay's LinkedIn is not a distribution channel pointing somewhere else; it is a destination, written to be read in the feed.
Cadence buys creative license. At 11.4 posts a week, a joke that fails costs nothing.
The obvious lesson, be funnier, is the wrong one. Reactions are not pipeline, and a company at Clay's scale can afford attention that does not convert, because the same feed carries the launches on a different day. A team publishing twice a week has no such license, because every post has to work.
One surface stays dark. Clay's founders post from personal accounts, and the LinkedIn APIs return company activity rather than individual feeds, so the founder-led half of the story sits outside what this method can see.
Google and LinkedIn are selling different companies
Clay's writing about its executive social program describes it as "all organic with no paid spend," and characterizes the paid element elsewhere as "a modest paid push, usually a thirty-second cut of a longer video dropped onto Instagram," where "the cost is tiny." On the same blog, Clay has published a post titled How Clay Uses Clay Ads: From $250 to $25 CPL — cost per lead — describing a continuously synced LinkedIn and Meta advertising operation built on enriched customer-record audiences.
Those describe different programs, so they are not in conflict. But Clay's public accounts emphasize the organic engine, and the ad libraries show a substantial, continuously refreshed paid one alongside it.
In the Google Ads Transparency Center, advertiser Clay Labs has run 118 creatives with a first appearance of January 18, 2024, split 59 text, 43 video, and 16 image. Activity is bursty: three months account for nearly half of all 118 creatives, with clear pushes in June 2025, November 2025, and June 2026.
new ad creatives
View the data
| Period | new ad creatives |
|---|---|
| 2024-01 | 3 |
| 2024-02 | 0 |
| 2024-03 | 0 |
| 2024-04 | 1 |
| 2024-05 | 0 |
| 2024-06 | 0 |
| 2024-07 | 0 |
| 2024-08 | 0 |
| 2024-09 | 2 |
| 2024-10 | 1 |
| 2024-11 | 0 |
| 2024-12 | 3 |
| 2025-01 | 0 |
| 2025-02 | 0 |
| 2025-03 | 0 |
| 2025-04 | 1 |
| 2025-05 | 6 |
| 2025-06 | 18 |
| 2025-07 | 4 |
| 2025-08 | 0 |
| 2025-09 | 0 |
| 2025-10 | 2 |
| 2025-11 | 19 |
| 2025-12 | 6 |
| 2026-01 | 1 |
| 2026-02 | 13 |
| 2026-03 | 6 |
| 2026-04 | 1 |
| 2026-05 | 3 |
| 2026-06 | 19 |
| 2026-07 | 9 |
Source: Google Ads Transparency Center, advertiser Clay Labs (AR02473231714533507073), all 118 creatives. Dated by firstShown. Captured 2026-08-03.
LinkedIn carries nearly four times the creative volume. Querying the LinkedIn Ad Library month by month across the year to August 3, 2026 returns 440 distinct creatives, with between 59 and 134 live in any given month. 2026 runs far hotter than 2025 — 250 new creatives in seven months against 58 in the final four months of 2025 — peaking at 62 in July.
new ad creatives
View the data
| Period | new ad creatives |
|---|---|
| 2025-09 | 5 |
| 2025-10 | 6 |
| 2025-11 | 10 |
| 2025-12 | 37 |
| 2026-01 | 5 |
| 2026-02 | 38 |
| 2026-03 | 59 |
| 2026-04 | 27 |
| 2026-05 | 32 |
| 2026-06 | 27 |
| 2026-07 | 62 |
Source: LinkedIn Ad Library, advertiser company 13018048, queried in monthly windows; each creative attributed to the first window it appears in. The August 2025 observation window is excluded because everything live at the start of observation would register as new. Captured 2026-08-03.
Volume is the easy half. Read the copy and the two programs turn out to have almost nothing to do with each other.
| Invites you to a live workshop or stream | LinkedIn 52% · Google 1% |
|---|---|
| Names a customer | LinkedIn 18% · Google 0% |
| Uses category or competitor keywords (best, alternative, top 10) | LinkedIn 18% · Google 65% |
| Uses database language (data, contacts, providers, records) | LinkedIn 55% · Google 73% |
| Asks for a demo or signup | LinkedIn 14% · Google 15% |
Both channels talk about data, which is unsurprising, since it is the product. The difference is what the data gets wrapped in. Half of LinkedIn creatives wrap it in an event and a fifth in a named customer; on Google, one in a hundred mentions an event and none names a customer. Google instead wraps it in category keywords at nearly four times the LinkedIn rate. The ask is the same on both, at about one in seven creatives. Everything before the ask is different.
On Google, Clay sells a database. The most-repeated headline, appearing 111 times, is Clay | Best GTM Data; the most-repeated description, 115 times, is "Give your sales teams a winning edge with the best data around." The rest reads like a category-keyword campaign: Best B2B Email Finder, Find Emails Instantly — Verified B2B Email Database, Track Buyer Intent Now. Claims are numeric and about supply: 850M contacts, 60M companies, 100+ data sources. Of 287 text variations, not one names a customer. The ads point almost entirely at two destinations, a dedicated /clay-for-gtm-ops landing page and the homepage.
On LinkedIn, Clay sells a calendar. The most-repeated headline, 28 times, is "Join the launch livestream for a first look!" After it come dated workshop invitations, each naming a session: a March 12 workshop on lead conversion, an April 2 workshop on account scoring, a June 3 workshop on re-engaging closed-lost data, a July 8 workshop on building an agent that preps reps for accounts. Copy opens by naming who it is for — "B2B Marketers —", "RevOps leaders -", "Marketers -" — then states a problem before it mentions a product. Customers appear constantly: Verkada, Rippling, OpenAI, Anthropic, Figma, Intercom.
Google captures demand that already exists. LinkedIn manufactures it, one workshop at a time.
The split makes sense once you read it as two jobs. Someone typing "best B2B email finder" into Google has already decided they need the category and is choosing a vendor, so the winning move is to claim supply. Nobody scrolling LinkedIn has decided anything, so the move is to teach them a workflow they did not know they wanted, in a room Clay hosts.
One detail cuts across both. Of the 440 LinkedIn creatives, 345 are sponsored status updates, paid amplification of posts that read like organic ones. The line between Clay's organic reach and its paid reach is therefore invisible from outside, and our LinkedIn engagement figures cannot separate the two.
The channel that looks dead and isn't
Clay's YouTube channel, @GrowWithClay, reports 16,500 subscribers and 762,940 lifetime views. Paginate the channel's Videos tab to exhaustion and you get 157 videos, the most recent published October 24, 2025. That is nine months of apparent silence.
The Videos tab excludes past live broadcasts. Pulling the channel's underlying upload playlist returns 163 videos, and the six it adds are all live streams, including one published July 24, 2026, ten days before our capture.
videos published
View the data
| Period | videos published |
|---|---|
| 2023-03 | 13 |
| 2023-04 | 1 |
| 2023-05 | 0 |
| 2023-06 | 45 |
| 2023-07 | 3 |
| 2023-08 | 2 |
| 2023-09 | 2 |
| 2023-10 | 4 |
| 2023-11 | 1 |
| 2023-12 | 2 |
| 2024-01 | 3 |
| 2024-02 | 7 |
| 2024-03 | 6 |
| 2024-04 | 8 |
| 2024-05 | 5 |
| 2024-06 | 9 |
| 2024-07 | 6 |
| 2024-08 | 5 |
| 2024-09 | 1 |
| 2024-10 | 1 |
| 2024-11 | 0 |
| 2024-12 | 2 |
| 2025-01 | 3 |
| 2025-02 | 23 |
| 2025-03 | 0 |
| 2025-04 | 4 |
| 2025-05 | 1 |
| 2025-06 | 0 |
| 2025-07 | 1 |
| 2025-08 | 0 |
| 2025-09 | 1 |
| 2025-10 | 2 |
| 2025-11 | 0 |
| 2025-12 | 0 |
| 2026-01 | 0 |
| 2026-02 | 0 |
| 2026-03 | 0 |
| 2026-04 | 0 |
| 2026-05 | 1 |
| 2026-06 | 0 |
| 2026-07 | 1 |
Source: ScrapeCreators YouTube video-details API, @GrowWithClay, all 163 uploads reconciled across the Videos tab (157), live-streams playlist (6) and Shorts playlist (0). Dates are exact ISO timestamps, not relative labels. Captured 2026-08-03.
By video count the channel does look finished. Count minutes instead.
minutes published
View the data
| Period | minutes published |
|---|---|
| 2023-03 | 86 |
| 2023-04 | 12 |
| 2023-05 | 0 |
| 2023-06 | 211 |
| 2023-07 | 16 |
| 2023-08 | 5 |
| 2023-09 | 24 |
| 2023-10 | 43 |
| 2023-11 | 8 |
| 2023-12 | 16 |
| 2024-01 | 20 |
| 2024-02 | 49 |
| 2024-03 | 35 |
| 2024-04 | 99 |
| 2024-05 | 33 |
| 2024-06 | 66 |
| 2024-07 | 34 |
| 2024-08 | 33 |
| 2024-09 | 7 |
| 2024-10 | 5 |
| 2024-11 | 0 |
| 2024-12 | 15 |
| 2025-01 | 41 |
| 2025-02 | 104 |
| 2025-03 | 0 |
| 2025-04 | 17 |
| 2025-05 | 9 |
| 2025-06 | 0 |
| 2025-07 | 160 |
| 2025-08 | 0 |
| 2025-09 | 61 |
| 2025-10 | 14 |
| 2025-11 | 0 |
| 2025-12 | 0 |
| 2026-01 | 0 |
| 2026-02 | 0 |
| 2026-03 | 0 |
| 2026-04 | 0 |
| 2026-05 | 231 |
| 2026-06 | 0 |
| 2026-07 | 231 |
Source: same 163 uploads, summing durationMs per publication month. One 2024 record is a live-stream placeholder that never broadcast and carries no duration; it is excluded. Captured 2026-08-03.
The two tallest bars in the channel's history are May 2026 and July 2026, the months that register as one video each. Both are Clay Cup, a live competitive tournament with 32 competitors and a $50,000 prize, broadcast in two single sittings of roughly 231 minutes.
Through October 2025, Clay published 161 videos averaging 7.6 minutes. Since then it has published two, averaging 231 minutes, thirty times longer.
By video count the channel died last October. By minutes published, its two biggest months ever came after that.
The transcripts explain what the channel was before it changed. Across 158 videos and 249,086 words, the dominant format is instruction. Forty-three titles begin "How to," and fifty-five either begin that way or carry the word tutorial, walkthrough, or guide. Twenty are numbered lessons in an explicit course, beginning at Clay 101 | Lesson 1: The FETE Framework and running as far as Lesson 23. All twenty are taught by the same narrator, opening with the same line: "hey it's Yash and in this lesson we're going to cover…"
The frameworks are named and proprietary: FETE, Jigsaw, and S.P.I.C.E. for prompt engineering. The most common phrases across a quarter of a million words are the vocabulary of a screen-share lesson: we're going, right here, you'll see, company domain, clay table. This is a curriculum with a syllabus, not a video marketing channel.
The second format is stranger, and barely mentions Clay. I sent 10,000,000 cold emails and learned this is the most-viewed video on the channel at 182,908 views. I Emailed 100 Celebrities. Here's Who Replied. I Tested Every Cold Email Script so You Don't Have To. I Let AI Train Me For a Marathon. These are creator-native formats, built on the first-person experiment and the audacious sample size, and they do audience acquisition rather than product education.
The channel had two jobs, then, and the tutorial work appears to have shifted to university.clay.com, where the archive shows 771 pages first appearing between April 2024 and March 2026, across the same period the lessons stopped. What remained on YouTube was the venue. A dashboard tracking videos per month would have shown a channel flatlining for nine months and prompted somebody to revive it. The channel was not flatlining. It had been repurposed.
The last six months, counted
For the six months from February 2026 to the capture date, here is everything we could count.
| LinkedIn organic | 11.4 posts per week (measured over a 41-day window) |
|---|---|
| LinkedIn paid | 245 new creatives; 63–105 live in any month |
| Google paid | 51 new creatives |
| 16 posts in this window; 18 since the account opened in January 2026 | |
| YouTube | 2 videos, 462 minutes — both Clay Cup broadcasts |
| X (main account) | Not retrievable via any endpoint we tried |
| TikTok, Threads | None found |
Instagram holds the widest gap in the dataset. The account has 18 posts total, and the median post drew 34.5 reactions. One took 108,715 reactions and 1,840,835 plays, more than three thousand times that median — a clip captioned "Maybe if I was using @claygtm like the rest of the team, we wouldn't be having this issue," carrying the tag #claypartner. The furthest-reaching artifact on Clay's newest channel is a creator collaboration rather than a Clay production, which fits the pattern everywhere else: the company is more often the host than the performer. It is also a reminder that a mean would describe this account badly.
The older X account, @clay_gtm, now called Clay University, runs from December 2018 to a final post on September 17, 2025 reading "We're launching a new account!!" The account it points to returns "account doesn't exist" from every endpoint we tried, so roughly eleven months of Clay's presence on X is dark to this method.
Publish the output, teach the work
Put the channels side by side and one organizing principle runs through them. With a single exception, Clay does not run campaigns about its product. It publishes the product's output, or it teaches the work the product does.
| Programmatic pages (15,173) | Publish the product output — enrichment answers, one page per company |
|---|---|
| YouTube (163 videos) | Teach the work — numbered lessons, named frameworks, creator-format demand content |
| LinkedIn ads (440) | Fill the classroom — dated workshops and launch streams, 52% of creatives |
| LinkedIn organic (67 in 41 days) | Hold the room — jokes for reach, dense launch posts for substance, almost no outbound links |
| Blog (238) | Maintain the evergreen library — 64% of sampled titles carry a year |
| Google ads (118) | The exception — a straight database pitch to people already shopping |
Search is the one surface where the audience arrives having already decided it needs the category, so teaching is wasted effort there and supply claims win. Everywhere Clay has to create the demand rather than harvest it, the format is a lesson, a workshop, or a worked example.
This is also why the community narrative is true but incomplete. Clay does have a community; whether it drives growth is exactly the sort of claim this method cannot test. What the retellings miss is that the community is fed by an industrialized teaching operation — a course on YouTube, a university subdomain, a workshop calendar behind more than half its LinkedIn ad creatives — and that the largest asset underneath all of it is a page generator producing the product's own answers ten thousand times over.
What actually transfers
None of this proves what made Clay grow; published output cannot. What follows is what a lean team can take from the shape of it.
Publish your product's output, not articles about your product. The /dossier build took about ten months and produces one flat page per company answering one question, the same question the product answers. It is a template, a data source, and a generator — closer to a data project than an editorial one. Most companies have an equivalent asset and never think to publish it, because it does not feel like content.
Sell differently on search than on social, deliberately. Clay's Google copy names no customers and claims supply; its LinkedIn copy names customers constantly and sells a seat at a workshop. Teams that run one message everywhere are using demand-capture copy on an audience with no demand yet.
A workshop calendar is an ad campaign. More than half of Clay's LinkedIn creatives point at a dated live session rather than a demo. That commitment is heavier than a creative refresh, because somebody has to run the workshop every time, and it is the part competitors are least likely to copy.
Volume, not bravery, is what makes a weird post possible. At two posts a week every post has to earn its slot, which is why most business-to-business social reads the way it does. The strange post is downstream of a cadence that makes any single post cheap.
Paid and organic run as one loop. 345 of Clay's 440 LinkedIn creatives are sponsored status updates. The organic post and the ad are the same artifact with different distribution. Teams that plan them as separate budgets, separate calendars, and separate creative pipelines are doing twice the work for less reach.
A channel can be repurposed instead of retired. Clay kept the YouTube audience and changed what the channel was for, moving the tutorial work to a different property. That is a harder move than stopping, and output-per-month reporting cannot see it.
The community story is the one Clay tells, and nothing here contradicts it. The story in the data is quieter: a company that decided marketing's job was to do the work in public, then built the machinery to do it ten thousand pages at a time.
How we did this research
Joule, Ampere's AI marketing agent, ran the collection across nine surfaces using the platform's market and competitive intelligence tools: clay.com and its archive record, the blog, LinkedIn organic, LinkedIn paid, Google paid, YouTube, Instagram, X, and a check for TikTok and Threads accounts. Page content came from a deterministic stratified sample of 186 pages across eight directories, and video content from transcripts of 158 of the 163 uploads. The collection took about two hours.
Raw responses were written to disk unmodified, and every figure was then computed from those files by a script rather than summarized by a model, so any number here traces back to a specific API response. The message percentages come from keyword rules applied to the full corpus: an ad counts as an event invitation if its text matches terms like live, workshop, webinar, register or join us, and as customer proof if it names one of fifteen companies Clay publicly cites. The rules are deliberately blunt, and they measure what the copy says rather than what anyone intended it to mean.
Reconciling took longer than collecting: checking blog dates against the archive, and rebuilding the YouTube history across three separate playlists before the shape of it was clear.
All figures captured 2026-08-03 from public sources: clay.com sitemap and pages, Internet Archive CDX, LinkedIn company and Ad Library APIs, Google Ads Transparency Center, YouTube, Instagram and X. Funding and valuation figures are from public reporting of Clay's August 2025 Series C and January 2026 tender offer. Clay's own statements are quoted from clay.com. We have made no claim about advertising spend, which is not publicly available. If Clay would like to correct anything here, we will update the piece and say what changed.