A midterm election is supposed to be the cheaper one. Two years ago the presidential cycle set the ceiling, and the off-year contests were expected to settle somewhere below it. That expectation has now failed. The forecast for the 2026 cycle sits above what the 2024 presidential race absorbed, and the money is arriving in channels that did not carry serious political weight in 2022.

The same pattern, in miniature, ran through almost everything that landed over the weekend. Agencies are moving budget into AI search before anyone has settled how to measure it. Meta's average ad price keeps rising while the complaints about what its feed carries keep accumulating. Google has quietly removed the one hard number its own documentation offered advertisers about when a bidding algorithm has finished calibrating. Microsoft is placing other companies' customer reviews underneath its ads. A survey of retail media buyers found 43% saying their own organisations underuse a channel whose impressions the industry is still arguing about how to count. A consultancy announced a vendor credential it had apparently held for seven weeks.

What connects them is not a shortage of numbers. It is that the numbers arriving fastest are the ones the counting party produced itself.

The midterm cycle passes the presidential one

AdImpact's projection puts total 2026 midterm political advertising at $11.6bn, a figure Digiday's Media Buying Briefing reported on October 5. The comparison that matters is not the last midterm but the last presidential year. The 2022 midterms drew $8.9bn. The 2024 presidential cycle drew $11.2bn. A midterm now outspends a presidential race, and does so by roughly $400m.

Jesse Contario, regional vice president for the southeast and political at MiQ, called it a record setting midterm. The description is literal rather than promotional: no previous off-year contest has approached this figure, and the normal rhythm of American political media buying has been that presidential years are the peak.

The channel split, which AdImpact issued in June, is where the structural change shows. Broadcast still takes the largest share at $5.6bn, slightly under half the total. Connected television takes $2.6bn, digital $1.6bn, cable $1.4bn. Radio draws $273m and satellite $88m. Set the $2.6bn of CTV against the $1.4bn of cable and the ordering has reversed from where it sat in 2022, when cable remained the second screen of political buying. By race type, Senate contests account for $3.4bn, gubernatorial races $2.4bn, House races $2bn, state legislature races $698m, and down-ballot issues and races $3bn. A further $20m is booked under US Government for spots announcing White House accomplishments.

Why broadcast still holds the plurality has less to do with preference than with friction elsewhere, and more to do with friction on broadcast itself. Todd Novick, vice president of political at AI Digital, listed rising rates, complications due to displacement, makegoods which nobody's a fan of, and lack of inventory. Political rates on broadcast are governed by Federal Communications Commission rules on lowest unit charge, which caps what stations may charge candidates but does nothing about the volume of competing demand in a tight market. Displacement and makegoods are the mechanical consequence: a spot bumped by a higher-priority buy has to be compensated with replacement inventory, often in a worse slot.

Audio is the fastest-moving line. Brian Conlan, president of Digital Audio Exchange DAX US, said political podcast ad spend is already up 42% year over year, with more expected in the final weeks before Election Day on November 3. That tracks a longer trend PPC Land has documented, with podcasts projected to take 48% of US digital audio ad spend by 2030 in VAB's modelling, and audio platforms adding transcript-level targeting through Comscore that lets buyers place against spoken content rather than show-level categories.

The targeting argument has shifted too. Jonathan Frohlinger, chief executive of Big Happy, argued that third-party targeting no longer decides elections, and that creative and location now carry the weight. Location in a political context means the designated market area and the precinct, not an audience segment. That framing fits what PPC Land reported when Gray Media moved its local inventory onto Madhive's DSP and when AdImpact added Polaris I/O signals to identify roughly 4,000 new advertisers a month.

Two prior findings sharpen the stakes. Adtaxi's June survey found social media had overtaken television as a source of political news at 29%, while VAB and Dynata found voters trusting television news roughly six times more than social media. Reach has moved to one place and credibility has stayed in another, which is a reasonable explanation for why $5.6bn still goes to broadcast while $2.6bn chases the same households through streaming apps.

Advertising Week opens on an admission

Advertising Week New York opened on October 5 at The Penn District in midtown Manhattan. Ruth Mortimer, the event's global president, summarised the mood in a sentence Digiday put in its headline: "We're all going to live with uncertainty forever."

That is a notable thing for a conference host to say at the start of a conference. The usual function of an opening address is to name a direction. Mortimer named a condition instead.

The programme underneath it is mostly about converting last year's conversations into contracts. Alistair Hill, chief executive and co-founder of On Device, described the shift as one toward implementation. Stephanie Latham, a vice president at Roblox, said the AI comfort conversations of a year ago had become enterprise agreements. Leslie Amadi, chief executive of Kochava, framed the week as a chance to see how far agentic AI has moved from talk to operations. PPC Land covered the same turn last year, when agentic AI infrastructure dominated the week's announcements without much in the way of signed business behind them.

The sharper change is who is in the building. Advertising Week expects 2,500 creators this year, up from 1,000 last year, a 150% increase in a single cycle. Dhar Mann holds a newly created role of chief creator officer. The stated target is $100m in creator deals over twelve months, which is the first time the event has attached a transaction figure to creator attendance rather than a headcount. Meta is running a Creator's Lounge on the second level. Mann is hosting a karaoke night. An omakase dinner with Rick Ross is scheduled for Wednesday. A leadership breakfast pairs Jenny Penich, chief revenue officer at Influencer, with Eleanor Hooker, head of community and customer marketing at Profound.

Adam Wescott, chief executive of Home Game Studios, put the commercial expectation plainly: "Creators don't just want the invite to help promote ticket sales. They want real deals to come from it." The $100m target is an answer to that complaint, and it is also a measurable commitment that can be checked in a year.

The measurement problem under the week's agenda is one PPC Land has reported repeatedly. AI-driven shopping is pulling budget out of search while creators seek credit for sales that complete somewhere other than the platform where the content ran. The ANA found 67% of marketers calling influencer measurement the hardest step in the discipline. Brands have responded by paying to treat creator content as a media asset, which converts an attribution question into a licensing one. IAB Australia found 85% of under-40s discovering brands through creators, a figure that explains the 2,500 badges without resolving who gets credited for the purchase.

Venue detail from the week reads like a map of who is paying for proximity. The Trade Desk opened at The Perch rooftop at PENN 2. Amex Ads took Riff Raff Club for a happy hour. The Landing at PENN 1 handles private dining, with Bar Primi on West 33rd Street and Little Rebel in the East Village hosting fringe events.

Agencies move search money before the measurement arrives

The Digiday Media Agency Report 2026, published October 5, surveyed 30 agency professionals during the third quarter of 2026 and convened a focus group of five senior media executives from Go Fish Digital, Horizon Media, Novus, Tinuiti and UM. The sample is small, and the report says so. What it captures is a direction of travel rather than a market size.

Two channels tied for the largest share of agencies increasing investment in 2026, both at 89%: AI search, and streaming video and connected television. Creator and influencer marketing and social media followed at 78% each. On the other side, 60% reduced out-of-home spending and 60% reduced broadcast. The figures are not net, which matters: 40% also reported decreases in social, in creator work, and in streaming video and CTV. Agencies are not moving in one direction. They are moving in both directions at once, within the same channels, which is what a market without an agreed measurement standard looks like.

Budgets themselves were close to flat. For 2026, 36% reported client budgets unchanged, 32% reported increases and 18% reported decreases. For 2027, 39% expect no change, 39% expect increases and 21% expect decreases. The money going into AI search is therefore not new money for most of these agencies. It is reallocated money, and the report traces where from: of the 77% planning to raise AI search and generative engine optimisation budgets within twelve months, 40% are pulling from general marketing and 20% from traditional search.

That last figure deserves its weight. One in five agencies raising generative search budgets is funding it by cutting paid or organic search. PPC Land reported a parallel shift in August, when AI search budgets surged while agentic ad tech moved on the buying layer, and again when 44.8% of brands said AI search was pushing them to spend more on public relations. Microsoft has published its own account of how AI search picks the brands it recommends, which is currently the most specific vendor description available of a process no third party can audit.

On AI costs, 45% said spending matched expectations, 31% found it higher and 24% lower. Among those reporting offsets, 70% cited efficiency savings and 17% cited replaced roles. Agencies are also charging for the capability: 59% are offering new products, 34% are selling AI strategy audits and 31% are adding service fees.

Creator work has become structural rather than experimental for this sample. Three quarters reported creator marketing becoming at least a little or significantly more important over the past year. The stated goals run community building at 61%, paid media amplification at 57%, branded content creation at 57%, authenticity at 50% and expertise or authority at 46%. In retail media, 17% of marketers partnered with influencers for content, second only to social at 96%.

Named contributors included Marcy Greenberger, chief investment officer at UM, Sarah Duffy, senior vice president for integrated investment at Horizon Media, David Dweck, president of Go Fish Digital, Olivier Pepin, chief technology officer at Novus, Harry Browne, vice president for TV, audio and display innovation at Tinuiti, and Stacey McCormick, chief marketing officer at Aerie. Monica Shukla, vice president of the biddable centre of excellence at Mile Marker, argued for fixing measurement definitions before a test begins rather than inheriting whatever metrics the platform surfaces afterwards.

Meta's ad prices rose while its reputation did not

PPC Land reported on October 4 on a BBC podcast episode of The Global Story, published September 30, in which presenter Tristan Redman interviewed Jeff Horwitz, the Reuters technology reporter who ran the 2021 Wall Street Journal investigation into Facebook's internal research. The episode ran 26 minutes across 16 chapters and was titled "Is Meta dying?".

Horwitz's central observation concerned an internal Meta marketing memo. Users, it recorded, absolutely thought the worst of Meta, and showed no decline in how often they used the product. From the company's perspective that pairing is not a problem to solve. It is the finding that makes the rest of the business model durable.

The commercial evidence sits in the quarterly numbers. Meta's second-quarter 2026 advertising revenue reached $59.36bn, 27% above the prior year. Impressions rose 14% and the average price per ad rose 12%. The Advantage+ automated suite passed a $75bn annual revenue run rate, and more than 9 million small businesses used at least one AI ad creative tool. Advertisers paid more per impression during a period in which the company's public standing was, by its own internal account, poor. PPC Land covered the same quarter from the earnings side, where profit fell 8% to $15.8bn as legal charges absorbed the ad gains, and the first quarter, where $56.3bn of revenue came alongside a doubling of advertiser adoption of AI tools. Advantage+ is now the default rather than an option for much of that spending.

Horwitz dated the mechanism to 2012. Before the IPO, Facebook was largely static pages with chronological friend updates and no real curation happening. After it, the company built a curated personalised feed optimised for time spent, volume of content produced and advertisement impressions. He described the resulting system as pro-acceleration: engagement with a content type produced more of that content type, whether the subject was cat videos, conspiracy theories, political rhetoric or weight-loss techniques.

Meta's internal integrity teams, assembled to catch rule-breakers and test healthier product changes, reached a conclusion Horwitz summarised bluntly. Optimising for growth, engagement and advertising revenue ran directly contrary to optimising for civil discourse. Sources passed him tens of thousands of blurry screenshots of internal work, photographed on personal phones.

Usage reporting has changed shape in the interval. Meta stopped publishing Facebook-specific daily and monthly active user counts after the fourth quarter of 2023, replacing them with Family Daily Active People across Facebook, Instagram, Messenger and WhatsApp. That metric averaged 3.60 billion in June 2026, 3% higher year on year, after dipping to 3.56 billion in the first quarter on internet disruptions in Iran and restrictions on WhatsApp in Russia. Instagram reports 2 billion daily active users and Threads 500 million monthly.

Publishers have felt the other side of the same optimisation. PPC Land reported in April that an unannounced feed change cut Digitalbox's Facebook reach by up to 75%, and in August that some advertisers lost placement controls with bid cuts capped at 90%. Trade bodies have acted where advertisers have not: IAB Sweden voted on March 11, 2026 to expel Meta over insufficient fraudulent advertising controls.

Capital allocation has moved to a different question entirely. Second-quarter capital expenditure hit $31.08bn against full-year guidance of $130bn to $145bn, and free cash flow fell to $784m from $8.55bn a year earlier. Headcount stood at 75,472 on June 30 after a May reduction of roughly 8,000 that generated $1.18bn in severance costs. Zuckerberg's 6,500-word AI strategy letter in August did not mention advertising once, which is a striking omission in a document from a company drawing $59.36bn of quarterly advertising revenue. Aaron Sorkin's film The Social Reckoning, with Jeremy Allen White playing Horwitz, is scheduled for US release on October 9 through Sony Pictures.

Google removes the 50-conversion number

Google has deleted the only hard threshold its documentation gave advertisers for when an automated bidding strategy has finished calibrating. PPC News Feed reported the change on October 4, crediting Hana Kobzova with spotting it.

The help page, titled "Duration of the learning period for campaigns and what affects it", previously told advertisers that calibration might require up to 50 conversion events or 3 cycles. It now says calibration typically takes 1 to 2 conversion cycles, with no conversion count at all.

The practical difference is substantial, because the two formulations are not interchangeable. A conversion count is a number an advertiser can read off a report and act on. A conversion cycle is defined by the campaign's own conversion lag, so a business with a two-day path to purchase and a business with a 30-day path are both told the same thing and get wildly different waiting periods out of it. The learning phase has always been the least observable part of automated bidding, and the 50-conversion figure was the one piece of it that behaved like a specification rather than a disposition.

The removal lands in the middle of a long sequence of bidding changes PPC Land has tracked through 2026. Google restored Target CPA and Target ROAS as standalone strategies after folding them into broader modes, shipped promotion mode alongside a wider bidding overhaul in August, and told advertisers that targets become the efficiency lever in budget-capped campaigns rather than ceilings. It also gave advertisers six weeks before CPA targets doubled under that change. Each of those moves made the relationship between an advertiser's input and the system's behaviour less direct. Removing the conversion threshold continues the pattern.

Google offered no explanation for the edit, and the article notes two readings: either the calibration process has genuinely improved to the point where a count is misleading, or the company has chosen to publish less prescriptive guidance. Neither can be distinguished from outside, because the signal that would settle it is internal. Conversion volume remains the input advertisers can influence through enhanced conversions and first-party data work, which Google itself quantifies through a Data Strength Uplift metric built on its own data strength scoring.

Microsoft puts Trustpilot quotes under its ads

Microsoft Advertising is showing one-line consumer review snippets beneath sponsored search listings, drawn from third-party sources including Trustpilot. PPC News Feed reported the format on October 3, crediting Khushal Bherwani with the find, with Barry Schwartz covering it at Search Engine Roundtable. The example captured showed a Best Buy ad carrying the line "They worked efficiently and swiftly".

The mechanic is a third-party endorsement placed inside paid inventory. Google has run comparable formats, and the structural question is the same in both cases: the advertiser buys the placement, a review platform supplies the sentence, and the search engine decides which sentence appears. None of those three parties is the one whose words are being quoted.

What is not disclosed is how the snippet is selected. A review extension drawn from a Trustpilot profile with a four-star average and a thousand reviews could legitimately surface either a glowing line or a lukewarm one, and the choice is consequential for click-through rate. Microsoft has not published selection criteria, refresh frequency, or whether advertisers can suppress specific quotes. Nor has it said whether the underlying rating is weighted or whether a minimum review count applies.

Microsoft has been expanding advertiser-facing controls through 2026 at a steady pace. PPC Land has reported the UTM fix separating Bing campaign types, exact match gaining total control over ad auctions, price extensions arriving in Microsoft Advertising Editor, ad customisers opening to all advertisers and a brand-safety control that blocks ads next to 1,000 page titles. The commercial backdrop is a search business whose advertising revenue passed $20bn as Copilot reshaped the result page.

A separate shopping test, reported on October 4 and credited to Sachin Patel, places sponsored shopping listings on a blue background. Background colour is the bluntest possible disclosure instrument, and it is worth noting that it is being tested at the same moment as a format designed to make paid listings carry the voice of unpaid reviewers.

Cold outreach gets deleted unread by 52%

Firmable surveyed 1,009 full-time professionals and found that 52% delete cold outreach without reading it. PPC Land reported the findings on October 2. A further 30% skim then discard, 11% read but rarely reply, 6% reply if relevant and 1% almost always reply. Chester Thompson, Firmable's vice president for the Americas, authored the report, published as The Cold Outreach Pulse Report and referred to elsewhere as the Cold Outreach Mood Report.

The volume context explains the reflex. Forty-four percent receive six or more cold messages weekly and 16% receive sixteen or more. Three quarters have blocked, muted or reported a sender.

The AI findings are the part worth isolating. Sixty-one percent distrust AI-written outreach they receive, and 67% claim they can spot and ignore it. Sixty-one percent find AI-generated personalised videos fake, against 7% who find them impressive. Meanwhile 33% of the same respondents use AI to help write their own cold or professional messages, and 54% send cold outreach themselves. Among senders, roughly 41% hold themselves to standards similar to those they expect of others and roughly 7% to lower standards. The gap between the receiving judgment and the sending behaviour is the survey's most durable result.

Rejection triggers were specific. Sixty-three percent reject messages that appear mass-sent, 37% reject generic greetings of the "Hi {FirstName}" variety, 36% reject a wrong name, role or company, 27% reject fake compliments and 20% reject stock phrases or excessive length. On what earns a reply, 58% cited relevance or genuine value, 26% named something genuinely useful on offer, 13% said human authorship mattered most and 10% cited visible research on the recipient.

The methodology is three sentences long, and PPC Land itemised what is missing: fieldwork dates, respondent countries, recruitment method, weighting, exact question wording, margin of error and subgroup sizes for the sector findings. Internal arithmetic does not close either. The report presents 22% and 4% as shares of senders while charting them as shares of all respondents, which recalculates to the 41% and 7% above. Generational sending figures of 28% for Gen Z, 21% for Gen X and 18% for millennials imply roughly a fifth engaged in cold outreach, against the headline 54%. Reply-driver options sum to 70%, leaving 30% unaccounted for. Respondents were 62% women and 38% men, 60% millennials, 20% Gen X, 17% Gen Z and 2% baby boomers.

Firmable's earlier work found B2B sales teams estimating 32% of their CRM data to be flawed, which is the upstream version of the same problem: a wrong name in a cold message is usually a wrong record in a database. Ahrefs separately found 7% of AI-using marketers publishing mostly unedited AI content.

Retail media buyers say they underuse the aisle

An IAB study commissioned by Grocery TV found 43% of US retail media buyers saying their own organisation underuses in-store retail media. PPC Land covered the research on October 2. The report, titled In-Store Retail Media's Full-Funnel Opportunity, surveyed 100 US retail media buyers and decision-makers in July 2026 through the IAB Insights Engine powered by Attest, covering brands, agencies and retail media networks.

The definition matters for reading the number. In-store retail media here means paid advertising in physical retail environments, including digital screens, audio and displays, and explicitly excludes traditional merchandising and trade promotions. A buyer saying the channel is underused is therefore not talking about shelf tags.

More than 90% rated awareness and reach, consideration, sales lift and loyalty as important or essential metrics. Fifty-nine percent put inventory and scale in their top three evaluation factors. On barriers, 27% cited activation complexity, 27% cited measurement limitations and 23% cited a lack of creative resources. Collin Colburn, IAB vice president of commerce and retail media, and Marlow Nickell, Grocery TV co-founder and chief executive, are the named spokespeople.

The most interesting finding was published without numbers. Organisations with dedicated retail media teams reported fuller utilisation than those where shopper or trade marketing controlled the budget. That is a claim about internal budget ownership determining channel adoption, and it arrived without the figures that would let anyone weigh it. PPC Land also noted the barrier percentages account for 77 of 100 responses, leaving the remaining 23 uncategorised, and that figures for measurement opportunities and CPM evaluation factors were withheld.

The measurement limitation that 27% named is not hypothetical. IAB Europe has been working through an in-store ad impression formula cut to six inputs, which is the standards body trying to fix a denominator before the inventory outruns it. Scale is arriving regardless: Walmart Connect runs 170,000 screens across more than 4,500 stores, Raley's Companies has a 208-store deployment, Perion acquired the PRN network, Amazon has added digital signage in grocery stores and McDonald's runs third-party ads on drive-thru boards at 450 US restaurants. Grocery TV itself added 3,500 audience attributes through an Esri data deal and published a separate survey finding 62% of shoppers buying after seeing an in-store ad.

The commissioning arrangement is disclosed and relevant. Grocery TV sells the inventory whose underuse the survey documents, and the full report was not among the materials available for review. A separate IAB Australia study found 56% of buyers keeping trade and media budgets apart, which is the structural condition the unnumbered finding points at.

A credential announced seven weeks after it appeared

Adswerve, a Denver consultancy, announced on September 29, 2026 that Adobe had granted it the Real-Time CDP specialisation. PPC Land reported the announcement on October 4 and found the credential already listed in Adobe's partner directory by August 9, 2026, spotted by the third party WISLR. Neither the release nor the directory explains the seven-week gap.

The specialisation has existed since May 19, 2023. Adobe's requirements, as stated in its May 2023 partner notice, call for three successful customer references, active membership at Silver, Gold or Platinum level, and staff certifications valid for at least 90 days from the request date. A global specialisation requires achievement across three of four regions: Americas, EMEA, Asia Pacific and Japan. Adswerve sits at Gold, one tier below Platinum, holds 71 Adobe certifications, and was one of 21 holders of the CDP credential at Gold or Platinum level in the August snapshot. Sara Hanrahan, principal lead for martech and an Adobe Champion based in Grand Rapids, Michigan, and Josh Stephens, also an Adobe Champion, are the named practitioners. Tom Zawacki became chief executive in February 2026. The firm acquired Analytics Pros in 2018 and merged the brand on January 1, 2019.

What the release does not say is the more useful half. No regional scope is given, so whether the specialisation covers one market or several is unstated. No client project sizes, outcomes or deployment counts appear beyond the phrase several enterprise deployments. Zawacki's claim that the platform delivers significant ROI for clients carries no supporting figure. On the agentic side, the main release says a human signs off before anything goes live while Hanrahan's LinkedIn post describes approval at every step, two materially different control models with no clarification of which governs. Nothing is disclosed about what gets logged when Adobe's CX Enterprise Coworker assembles an audience, or who carries responsibility when an approved plan underperforms. A Journey Optimizer specialisation is implied as in progress without a date, and the reference to key ecosystem partners names none of them.

The underlying market is real enough. Adobe has shipped a data collaboration tool for privacy-first advertising, added cross-border footfall attribution to its DSP, live first in APAC, and supplied AI tools to all 32 NHL clubs for fan personalisation. A partner credential is a claim about capability made by a vendor about a reseller of that vendor's software, which is a narrower kind of evidence than a client outcome, and the distinction is the point. The martechcertification economy runs on exactly this sort of attestation.

Also noted

  • October 4 - Bing is testing a blue background behind sponsored listings in shopping search results, a change spotted by Sachin Patel that uses colour rather than labelling to separate paid results from organic ones. PPC News Feed
  • October 3 - Google is testing two separate dotted menus on a single search ad, one beside the URL and one beside the headline, each opening either the landing page or a "Why this ad" panel, observed in car detailing listings in Brisbane by Khushal Bherwani. PPC News Feed
  • October 3 - Google's "Destination not working" policy help article, updated on September 30, now tells advertisers to configure at least one asset group to avoid an invalid final URL error during campaign setup, with the company stating that enforcement procedure is unchanged. PPC News Feed
  • October 2 - The Trump administration has spent nearly $2.9m on three spots framed as public service announcements ahead of the midterms, with Federal Communications Commission chair Brendan Carr defending the classification, in a roundup that also recorded Accuracast finding OpenAI reporting 100 clicks where independent tools counted 20. AdExchanger
  • October 2 - Fox News took September 2026 primetime with 1.717 million total viewers against MS NOW's 839,000 and CNN's 539,000, while in the 25-54 demographic CNN's 90,000 sat just 11,000 ahead of MS NOW's 79,000 as all three networks fell month on month. Adweek