Publishers among the 50,000 largest US websites added a net 54,942 ads.txt lines in the period covered by DataBeat's September 2026 Sellers Report, published on September 18, 2026, and 44,608 of those lines came through reseller relationships - more than four times the 10,334 added through direct ones. The result undoes the reseller pruning the same tracker recorded a month earlier.

In Short

Big US websites keep public lists of the companies allowed to sell their ad space, and in the latest monthly count they added far more middlemen to those lists than companies they deal with directly. That matters to advertisers because every extra middleman is another route to the same ad slot, which can mean paying more fees or ending up bidding against yourself. A month after publishers were trimming those middlemen, the trimming has stopped, at least for now.

A report, and a newsletter

DataBeat, a programmatic data business that operates under MediaMint, posted the September edition of its Sellers Report on its website on September 18. MediaMint's DataBeat newsletter carried the headline figures to subscribers today, wrapped in commentary on agentic advertising, CTV operations and a sales pitch for a yield-management product. The report carries the data; the newsletter adds interpretation and, in places, a commercial agenda.

The dataset is narrow by design. According to DataBeat, the report analyses the ads.txt files of the top 50,000 US publishers and tracks how the lines inside them change over roughly a month. Publishers are sorted by traffic rank into three bands: High-Traffic (ranks 1 to 500), Mid-Traffic (501 to 2,000) and Low-Traffic (2,001 to 50,000). The exchanges selling that inventory are ranked separately, by their number of direct connections, into Established (1 to 50), Scaling (51 to 150) and Emerging (151 to 500).

Net growth sits on top of heavy churn

An ads.txt file is a plain-text list, hosted on a publisher's own domain, naming every company authorised to sell that publisher's inventory. Each line carries a field declaring the relationship as DIRECT, meaning the publisher holds the account with that exchange, or RESELLER, meaning an intermediary is authorised to sell through someone else's account. It records permission, not volume.

According to DataBeat, publishers in the sample added 459,733 lines during the period and removed 404,791. Direct entries accounted for 191,538 additions and 181,204 removals, a net of 10,334. Reseller entries contributed 268,195 additions against 223,587 removals, a net of 44,608. On that basis resellers delivered 81% of the market's net growth and 4.3 times the direct figure.

The ratio looks different from the gross side. Reseller lines made up 58.3% of all additions, barely above the 57.4% share of gross additions that DataBeat recorded in its June 2026 edition. What moved was the removal column. Publishers deleted 94.6% as many direct lines as they created, against 83.4% for reseller lines, so direct relationships churned almost as fast as they were written. Additions and removals together came to 864,524 line changes, which means the headline net figure represents roughly 6% of all activity in the files.

Set against June, when the market netted 60,398 lines and the direct segment 21,300, the September direct net has roughly halved. The reseller net implied by June's totals was 39,098; September's 44,608 is about 14% higher.

August's pruning, reversed

According to DataBeat, the September net marks "a sharp reversal" from the previous edition, which recorded a near-zero net change in a month when publishers cut reseller paths while protecting direct ones. That edition was published on August 6, according to DataBeat's list of posts.

DataBeat's own summary is blunt. A single month of reseller-led growth, it wrote, does not undo months of supply chain discipline, but "shows that discipline is not permanent." The report attributes the shift to buyers "appearing to seek more reach this month" - an interpretation rather than a measurement, since the report contains no bid, spend or clearing-price data to support it.

The longer series is uneven. DataBeat's July 2026 edition recorded a net addition of 64,800 lines, with resellers accounting for 56.9%, and April's data put the reseller share at 63% of net gains. Those percentages were not all calculated on the same base: the June figure described gross additions, while April's was a share of net gains. Mixing the two bases distorts the trend.

Nativo leads the SSP table

According to DataBeat, Nativo added 9,814 lines, the most of any supply-side platform, or SSP, in the sample and its strongest month yet. The report's tables label the company "Life360 Ads Featuring Nativo." Life360 agreed on November 10, 2025 to acquire Nativo for approximately $120 million in cash and stock, according to Life360, which said at the time it expected the deal to close in January 2026. DataBeat credits part of Nativo's momentum to Life360 scaling its advertising platform.

The report ties Nativo's gains to its targeting design: an ad platform built on continuous location and household-level signals rather than third-party cookies, with existing integrations to demand-side platforms that give publishers a route to buyer demand without separate cookie-based targeting infrastructure. Risecodes, which appears in the tables under its RISE brand, ranked second at 6,856 lines. DataBeat attributes that gain to Risecodes' AI and machine-learning-based Smart Auction Management product. Media.net (4,985), Sharethrough (4,654) and InMobi (4,088) completed the top five.

Together, those five companies added 30,397 lines - 55% of the net market total.

Tier by tier

All three publisher bands turned positive. Low-Traffic sites added 296,394 lines and removed 261,656, a net of 34,738. Mid-Traffic sites netted 10,522 (105,059 added, 94,537 removed) and High-Traffic sites 9,682 (58,280 added, 48,598 removed). The long tail dominates both sides of the ledger: the Low-Traffic band, 48,000 of the 50,000 sites, produced 64.5% of all additions and 64.6% of all removals. According to DataBeat, the previous month's cleanup had hit High-Traffic sites hardest; every band is now adding more than it removes.

The leaders differ by band. Among High-Traffic sites, PubMatic led at 463, followed by InMobi (416), Sharethrough (400), Rubicon (372) and Nexxen (353). The Mid-Traffic table was headed by RISE (906), Nativo (883), InMobi (863), Media.net (732) and Sharethrough (624). In the Low-Traffic band, Nativo added 8,771, ahead of RISE (5,747), Media.net (3,908), Sharethrough (3,630) and InMobi (2,809).

InMobi and Sharethrough are the only companies to appear in the top five of all three bands, according to DataBeat. Their per-band figures sum exactly to their overall totals - 416, 863 and 2,809 make 4,088 for InMobi; 400, 624 and 3,630 make 4,654 for Sharethrough - which is a useful check that the tables are internally consistent.

DataBeat reads the High-Traffic table as evidence of "premium publishers' continued preference for established exchange infrastructure," pointing to PubMatic and Nexxen, which appear in no other band's top five. So does Rubicon, the name under which Magnite's exchange historically operated, at 372 lines. The commentary does not mention it.

Where commentary and arithmetic part ways

One claim in the report does not survive a check against its own numbers. According to DataBeat, "Low-Traffic's growth is being driven almost entirely by one SSP's expansion." Nativo's 8,771 Low-Traffic lines are 25% of that band's 34,738 net. The four companies behind it added 16,094 between them, and the top five together account for 72% of the band's growth. Nativo is the largest single contributor by a clear distance. "Almost entirely" overstates it.

Direct connections: BRAVE, RichAudience and an unexplained jump

A second set of tables counts publisher domains on which an SSP gained a direct connection, rather than lines. According to DataBeat, RichAudience led the Established group with 126 new direct domains, which the report says was likely helped by its natural-language-processing-based contextual targeting. Nativo followed at 84. Minute Media (33) and TrustedStack (24) also appear, alongside an entry at 61 identified only by a pair of logos.

In the Scaling group, BRAVE added 509 domains, which DataBeat describes as leading "by a margin." The margin is 24 domains: Adipolo added 485. iMDS (290), Copper6 (185) and PMC (119) followed. The report links BRAVE's gain to recent server and identity infrastructure upgrades built for direct publisher scale.

The Emerging group was topped by Global Sun Media at 169 domains, then Viously (155), Bidgency (120), Freestar (114) and Hashtag Labs (104). DataBeat is candid about the leader: the jump has "no clear public explanation," and the report treats it as a strong onboarding month rather than a strategic shift.

Rank movements, based on direct coverage among the top 50,000 publishers, tell a similar story. In the Established group, InMobi and Minute Media each rose three places, Ogury and Zeta two, and OpenX one. In Scaling, BRAVE climbed 19 places, iMDS 15, Adipolo and IncrementX seven each, and Kueez four. The largest moves came in Emerging, where Global Sun Media rose 87 places, Sellwild 80, BidsCube 67, Liqwid 36 and Viously 30.

DataBeat finds a common thread among the direct-connection winners: cookieless-ready targeting through contextual signals, location data or scalable infrastructure. That framing arrives more than a year after Google said it would keep third-party cookies in Chrome, ending a phase-out effort that began in August 2019. The report offers no evidence that buyers chose these SSPs for their targeting. The causal language - "likely helped," "backed by," "drawing on" - is DataBeat's inference.

Duplication edges up in the middle tier

The report's most consequential numbers for buyers concern auction duplication. According to DataBeat, Scaling and Emerging SSPs rely on Established platforms as demand partners, and in the process Established SSPs can end up listed as a reseller on domains where they already hold a direct connection. The same publisher inventory then reaches the same exchange by two routes, and a buyer can find itself bidding twice on one impression. The problem is structural to the supply path that most open-web impressions travel.

DataBeat measures this in two ways. The first splits each tier's domains into Direct Only, Reseller Only and Direct + Reseller. For Tier 1 SSPs, the September split was 19.66% direct only, 44.72% direct plus reseller and 35.62% reseller only. Tier 2 stood at 26.77%, 38.06% and 35.18%; Tier 3 at 32.22%, 35.08% and 32.70%.

The second measure counts a domain as duplicated when it connects to an SSP through more than one intermediary or integration. On that basis, 46% of Tier 1 domains were duplicated, with an average intermediary count of 1.33. Tier 2 reached 39% (1.20) and Tier 3 36% (1.19).

Compared with June's figures - 46%, 38% and 34%, with average intermediary counts of 1.31, 1.20 and 1.18 - Tier 1 has not moved on the headline rate, while its intermediary count has crept up. Tier 2 is one point higher, which DataBeat flags as "a small move worth watching." Tier 3 has risen two points since June and its direct-only share has fallen from 35.15% to 32.22%, a shift the report's commentary does not address.

DataBeat's explanation for the Tier 2 increase is mechanical: as Scaling-tier SSPs add direct publisher relationships at pace, some of that growth lands on domains where a reseller path already exists. "Direct connection growth and clean supply paths are not the same thing," according to DataBeat.

DataBeat also sells a Competitive Intelligence tool to help SSPs cut ineffective intermediaries and duplicate domains - the same dashboard, according to DataBeat, that underpins the Sellers Report. The firm measuring the problem sells tools to address it.

What the method cannot show

The report covers US publishers only, measures change over "the past month" without stating crawl dates, and ranks SSPs by direct-connection counts rather than by revenue. The SSP line totals are presented as contributions to net growth, though the report does not publish SSP-level removals alongside them. A publisher can also list an SSP and never sell a single impression through it.

None of that makes the data unimportant. Ads.txt remains the only public, machine-readable record of which sellers each publisher authorises. But a line count is a proxy. It cannot establish that buyers are spending more through resellers, only that publishers have given resellers more permission to try.

The newsletter around the report

The newsletter's commentary on third-party coverage was notably sceptical. On advertising within AI agents, MediaMint argued that the performance metrics come from the companies selling the ad stack and are not neutral, and it favoured keeping budgets in test mode until independent verification exists. That caution sits alongside DataBeat's own data from the summer, which found AI agents took part in 86% fewer auctions than conventional buyers.

On Magnite's ambitions to control the auction's decisioning layer, MediaMint warned that "buyers could end up paying for the same decision twice," once through Magnite and once through their own DSP. Magnite set out two routes for running partner AI models inside its auction in a position paper in early August, while its second-quarter results, released on August 5, attributed no revenue to agentic or containerised products.

On CTV, MediaMint urged care over a study finding supply path issues in more than half of CTV programmatic services, noting that open exchange is a small part of CTV spend. The point echoes an IAB study reported in July, in which 43% of CTV buyers doubted where their ads ran and confidence in open-exchange purchases fell to 33%.

A separate section promoted Mia Yield, a MediaMint product. Under the headline "The Five Blind Spots Costing Publishers 10-15% of Yield," MediaMint argued that auctions re-price every hour while most publishers review their yield stack weekly or monthly. A price floor set too high leaves no log, a demand partner drifting off pace sends no alert, and an ad that fails to render simply does not count. Mia Yield, according to MediaMint, monitors the stack continuously, applies routine fixes within limits the customer sets, flags exceptions and logs its actions.

The performance claims come with no published methodology. MediaMint said its own teams now spend about half the time they used to on manual reviews, that "they hold accuracy at 100%," and that a 15-20% uplift follows. It did not define accuracy, name clients or describe a control group. The two ranges also sit awkwardly together. Recovering a 10-15% revenue leak would produce an uplift of roughly 11-18% on the depressed base; a 20% uplift implies a leak of nearly 17%, beyond the range in the headline. These are vendor figures and nothing more.

Why the direction matters

Why would a month's worth of text-file edits matter to anyone spending money? Because each reseller path carries its own fee. IAB Spain's first technical guide to SSPs estimated that a direct SSP connection can pass 70-80% of advertiser spend to the publisher, against 40-50% for indirect paths involving multiple resellers. Those figures come from Spain, not the US. Still, a month in which 81% of net ads.txt growth ran through reseller entries tilts the available routes toward the more expensive end.

The structure of the file itself is also under pressure. On March 30, 2026, Scope3 chief executive Brian O'Kelley proposed adagents.json as a structured replacement for ads.txt, adding delegation types, placement identifiers and geographic scoping. DataBeat's entire analysis rests on a two-value field - DIRECT or RESELLER - that such a format would replace with finer categories. Whether that would make duplication easier to police, or simply harder to count, remains open.

For now, the September data points in one direction. Publishers pruned intermediaries in August, then added them back as soon as, in DataBeat's reading, buyer appetite for reach returned. With Q4 buying about to begin and Tier 1 duplication stuck at 46% for at least four consecutive editions, the gap between supply growth and supply hygiene is not narrowing.

Timeline

Summary

Who: DataBeat, a programmatic data business operating under MediaMint, and the SSPs and publishers in its sample, led by Nativo (now marketed with Life360), Risecodes, Media.net, Sharethrough and InMobi.

What: DataBeat's September 2026 Sellers Report found publishers among the top 50,000 US sites added a net 54,942 ads.txt lines, 44,608 of them through reseller relationships against 10,334 direct, reversing the previous month's reseller cuts. Tier 1 SSP duplication held at 46% while Tier 2 rose to 39% from 38%.

When: The report was published on September 18, 2026. MediaMint's DataBeat newsletter circulated the findings today.

Where: The data covers ads.txt files of the top 50,000 US publishers by traffic rank.

Why: Reseller paths add fees and create duplicate routes to the same impression, so a month in which publishers add intermediaries faster than direct relationships affects how much of a buyer's spend reaches publishers and how often buyers bid against themselves. DataBeat attributes the shift to buyer demand for reach, though the report contains no spend data to confirm it.