
ChatGPT Ads reached Poland with the personalization switched off.
The 31 markets that went live on August 24 are exactly the EEA plus Switzerland, the territory OpenAI carved out of personalization. Polish buyers get a contextual product with a 65 PLN daily floor and a CPC figure that is a suggested bid, not a market rate.
This week’s picks
Editorial channels, one operator perspective.
AI & Data
Targeting, attribution, the data behind AI ads. The new fundamentals.
Browse channelPPC
Performance Max, Advantage+, paid social. For people who run the campaigns.
Browse channelSEO
Core updates, AI search, technical SEO. What’s actually moving rankings.
Browse channelMarTech
Stacks that earn their keep, and the ones quietly subsidized by venture money.
Browse channelE-Commerce
Conversion, checkout, the AI shopping behavior reshaping retail funnels.
Browse channelHow-To Guides
Step-by-step playbooks you can run on Monday. Tested in real accounts.
Browse channelData & Research
Original analysis and the numbers behind the trends: sources shown, claims checked.
Browse channelDigital Marketing
The cross-channel picture: strategy, maturity, and the tool stack that ties it together.
Browse channelCRO
Conversion rate optimization: landing pages, checkout, and the experiments that move revenue.
Browse channelRecent articles
Most audit findings are assertions. These six leave evidence.
The paid media audit is sold on percentages, and chased to their sources most end at another agency blog. Here are six failures you can prove inside your own account in an afternoon, and the arithmetic that separates a finding from an assertion.
ACP and UCP are converging. The merchant work underneath was never the argument.
Roundups frame agentic commerce as a standards war between OpenAI and Google. Read both specifications and OpenAI’s retreat from Instant Checkout, and the two rails are converging on one journey, asking merchants for the same feed at a cadence the spec never demanded.
Google’s 7% and the independent 16% never measured the same advertisers.
Two migration dates are circulating with the wrong campaign types attached. Google’s 7% and the independent 16% come from different verticals, baselines, and years, and the question worth testing is whether the added volume was ever new.
The watermark counts the words Claude chose. It does not say who wrote the page.
Claude’s newest models weave an invisible mark into their text, with no documented opt-out. In Anthropic’s explanation the mark measures one thing, how many words the model chose: faint on a proofread draft, loud on prose it wrote end to end.
ChatGPT Ads reached Poland with the personalization switched off.
The 31 markets that went live on August 24 are exactly the EEA plus Switzerland, the territory OpenAI carved out of personalization. Polish buyers get a contextual product with a 65 PLN daily floor and a CPC figure that is a suggested bid, not a market rate.
The 30-conversion rule is real. It just answers a different question.
Teams running several European markets argue about splitting campaigns per country, and settle it with a number somebody half-remembers. Google publishes it as the sample you need to read a result, not the volume the algorithm needs, and that changes the decision.
The thresholds you restructure your account around were never published.
Thirty conversions a month. Fifty events a week. Every consolidation argument in the industry rests on those numbers, and neither platform states them as structural requirements. Google says something quieter and more useful, and once you read what it actually says, the real cost of splitting a campaign turns out to be arithmetic you can do yourself.
You put AI into the workflow. Your quality control is still a feeling.
Every team now has models writing, classifying, and summarizing at volume, and almost none can say whether the output got better or worse last month. The fix is an eval: a frozen set of real inputs, a rubric that discriminates, and a threshold. The research on how to build one is unusually clear, and unusually uncomfortable about the shortcut everyone takes.
Everyone reports AI traffic from a channel grouping that does not exist.
The slide says AI traffic is up and to the right. Underneath it, some assistants send a referrer and some send nothing, the biggest AI surface of all arrives labeled as ordinary organic search, and the crawler that decides whether you get cited at all is a different bot from the one everybody blocked. Here is what is actually measurable, and what to stop claiming.
Most creative tests end long before they could have concluded.
Two things break creative testing at once: the platform decides who sees what, and almost nobody runs the numbers on how much data a real answer needs. The arithmetic is not close. Detecting a 20% difference takes roughly 400 conversions per variant, and most tests are called at a tenth of that.
The platform buys whatever you last told it about. For most B2B, that is a form.
Your sales team knows which leads were worthless. Your ad account does not, because nothing ever told it. Closing that loop is not a reporting project, it is a change to the optimization target, and the hard parts are a seven-day window, a hashing spec the two platforms disagree on, and the double-counting trap nobody warns you about.
Average ROAS stays healthy long after the next dollar stops paying.
The number on your dashboard divides all the revenue by all the spend, which describes money you already committed. The decision in front of you is about the next increment, and that return is always lower than the average and falling. Here is how to find where your curve bends, and what to do at the point the two numbers disagree.
The feed is the targeting. Everything else is a thin wrapper.
In Shopping and Performance Max you cannot pick keywords, audiences, or placements in any way that sticks. What decides which auctions you enter is a text file: the title, the attributes, and the labels you put on each product. Here is how the matching actually works, and the one Google rule about AI-written titles that almost nobody has implemented.
Your bidding is automated. The number you feed it is not.
Google and Meta will optimize toward whatever a conversion is worth, and they take your word for it. Most accounts send the order total, which is not profit, then set a target on top of a number the business never earned. Here is what to send instead, and the refund pipeline that keeps it honest.
Connect Google Analytics to an AI assistant, and let it do the reporting.
The whole path, in order: the Google Cloud project, the credential, the connection, and then the part that actually pays for itself. One scheduled pull becomes a dashboard, a cohort grid the interface cannot draw, a written weekly summary, and a Slack alert that stays quiet until something real moves.
Labeling your AI creative is the law now. That it costs you sales is a rumor.
From August 2, three separate regimes (the EU AI Act, New York’s synthetic-performer law, and Amazon’s marketplace policy) start forcing you to disclose AI-generated creative. The panic that the label tanks conversion is repeated everywhere and proven nowhere. Here is what each rule actually covers, and how to decide where a human face still earns its cost.
Build a Google Ads agent that can’t wreck your account.
You can run Google Ads from a terminal now: read the account, draft the change, spend the money, all in plain English. This is the build guide. Which of the four paths to use (scripts, the official read-only Model Context Protocol (MCP) server, a community write server, or the raw v25 API), what v25 quietly broke, and how to wire a write path that fails safe instead of fast.
Don’t panic-bid on your own brand after the spam update.
The June 2026 spam update set off a wave of “brands lost their own results” panic. No primary data supports it. What actually got demoted was the coupon and affiliate sludge on your brand-plus-modifier queries, and branded search is the one segment AI Overviews are making stronger.
I shipped an AI iOS app solo in 30 days. Writing the code was the easy part.
Tymbre went from a clickable mockup to the App Store in 30 days, built by one person after hours, with AI. Here is the full product: 1,116 launch-period commits, a self-hosted core architecture in Warsaw, adversarial QA, and the lesson that the bottleneck was never the typing.
ADAM: four pillars, five levels, and the one you’re quietly starving
AI, Digital, Automation, Measurement. Every business already runs on all four, and is fragile in the exact one nobody is watching. ADAM is a diagnostic that plots you on a 4×5 matrix, names your weakest pillar, and forces one decision per quarter. Here is the whole framework, and the free playbook.
Claude Code can operate your ad accounts. The question is what you let it touch.
Claude Code already pulls last week’s numbers, drafts the negatives, and stages a budget change across Google Ads and Meta, wired in through the Model Context Protocol (MCP) and the platforms’ own APIs. Here is how the plumbing works, why Google’s own connector refuses to spend a cent, and where the guardrails still have to come from you.
Model Context Protocol: one plug for every AI marketing tool
Model Context Protocol (MCP) turns a large language model (LLM) from smart autocomplete into an agent that can touch your CRM (customer relationship management), analytics, and ad platforms through a single standard. How the handshake works, who adopted it, and the security perimeter it opens.
Marketing’s AI reorg is real. The layoffs aren’t.
Seventeen percent of chief marketing officers (CMOs) have cut marketing jobs because of AI. Thirty-six percent expect to within two years. The gap between those numbers is where the restructuring is actually happening.
One agent will not run your marketing. A team of them might.
When a single AI agent hits its ceiling, the work moves to a fleet with an orchestrator on top. The 2026 problem is no longer model quality. It is coordination, guardrails, and reliability.
Half your clicks are gone. Your best ones aren’t.
AI Overviews sit on nearly half of Google queries by most 2026 counts, and have cut clicks to the top organic result by 58%. The traffic that survives is disproportionately the traffic that was going to convert.
Your best-performing channel might just be a passenger
Last-click and even a mix model can flatter the spend that only harvested demand it never created. Incrementality testing is the causal tie-breaker: geo experiments and holdout lift tests that show which dollar actually caused a sale.
Performance Max isn’t a black box anymore. You’re just running it like one.
Search themes, campaign-level negatives, brand exclusions, and channel reporting quietly turned Google’s most opaque campaign type into a controllable broad-match machine. The controls exist. Most accounts still don’t use them.
Google Analytics 4 measures behavior well. It was never your system of record.
Google Analytics 4 (GA4) got good at engagement, e-commerce, and prediction, and stayed bad at pretending to be a database. In 2026 the working setup reads the interface as a privacy-filtered view and moves the source of truth to BigQuery.
Two companies own retail media. Everyone else is a test budget.
US retail media hits roughly $71 billion in 2026, but Amazon and Walmart take nine of every ten new dollars. The channel is huge, growing fast, and more concentrated than any other in digital.
Do you actually need a CDP in 2026, or is the warehouse enough?
Packaged customer data platforms and warehouse-native, composable stacks now do most of the same jobs. The real fork is not buy vs build. It is where you put identity and consent, and how fast you need to act on them.
Advantage+ reports one ROAS. It’s the wrong one to trust.
Meta made Advantage+ the default and lowered the bar to use it. The performance lift is real. But the number it reports is blended, and the control you give up is not. Here’s how to run it without fooling yourself.
Your AI strategy is an operating-model problem, not a tool list
In 2026 the marketing teams pulling ahead with AI are not the ones with the most tools. They are the ones that redesigned how the work is done, who decides, and how the payback gets proven.
The referral fee is the smallest line on your Amazon bill
Amazon called its 2026 fee change slight, and for one unit it is. The real cost of selling on Amazon and Walmart sits underneath it: a stack of fulfillment, storage, and ad charges that turns a 6-15% referral into something operators put near 30-40% of gross merchandise value (GMV), and each platform hurts in a different place.
Outdoor screens grew up: programmatic DOOH is now audience media
In 2026 the billboard reports back. Programmatic buys the majority of digital out-of-home, measurement moved from impressions to incrementality, and outdoor screens now compete for performance budgets, not just awareness lines.
People watch connected TV. Budgets haven’t caught up. That’s the trade.
Streaming is now nearly half of US TV viewing, but connected TV still draws only about 8% of ad spend. The gap is the last piece of underpriced reach in video, if you can survive the measurement and frequency chaos to claim it.
The blast is dead. Email in 2026 runs on lifecycle and deliverability.
Gmail and Yahoo turned deliverability into the price of entry, and behavior-led flows now compound where broadcast sends stall. What actually moves email revenue in 2026, and what to distrust in the return-on-investment decks.
How to get quoted by the machine that replaced your search traffic
Generative engine optimization isn’t SEO with a new coat of paint. It’s writing for extraction (answer-first, evidence-dense, cleanly structured) so ChatGPT, AI Overviews, and Perplexity pick you as the source. This piece is built the way it tells you to build.
Your affiliate program’s real cost isn’t the commission rate. It’s the incrementality.
Pay 10% on every affiliate sale and a chunk of those buyers were coming anyway. The 2026 shift is away from coupon volume toward deeper partnerships, but the number that decides whether the program pays is the one nobody prints on the invoice.
You’re not measuring a third of your conversions. Here’s the fix and its price.
Client-side tracking now loses 20-40% of events, more in Safari-heavy markets. Server-side tagging recovers a real chunk of it, but it’s infrastructure with a bill and a consent trap, not a plugin you switch on.
On June 15, one switch decides whether your conversions still exist.
Google Analytics stops reading its own ad settings and hands the entire decision to Consent Mode. If your consent management platform (CMP) wires the wrong signal, EU conversions and remarketing quietly fall off, and Analytics will still look fine.
The ad-ops agent is real. Full autonomy is a story you’re being sold.
AI agents already run bidding, budget shifts, and creative rotation end to end. But 2026 is the year of assisted autonomy: humans on the guardrails, not off the loop. The gap between those two things is where accounts get wrecked or saved.
Half your display dollar never reaches the publisher. Here is where it goes.
Independent audits keep finding the same thing: only about half of a programmatic display dollar reaches the site that ran the ad. The rest is the ad tech tax, and in 2026 the way to claw it back is structural, not another blocklist.
The average LinkedIn CPC is a number that will lose you money.
Published 2026 benchmarks put LinkedIn cost per click (CPC) anywhere from $5.58 to $10.11, and as high as $15.72 in peak quarters, with cost per lead (CPL) from $15 to $350. The spread isn’t noise. It’s the whole game. Plan to the range that matches your audience, not the headline that matches nobody’s.
Automate the plumbing, not the judgment: n8n and Make
The 2026 workflow-automation choice is not really n8n versus Make. It is knowing which marketing tasks belong to a no-code canvas, which need real code, and which should stay in a human’s hands.
Marketing mix modeling is back. Not to replace attribution, to overrule it.
Privacy killed the click-level view, and marketing mix modeling (MMM) returned to fill the hole. But the brands getting it right don’t run MMM instead of attribution. They run three layers, each doing the one job it’s actually good at.
RAG for marketing: how to ground an LLM in your own brand knowledge
Retrieval-augmented generation is how you stop a large language model (LLM) from inventing your product facts. In 2026 the work is a governed pipeline, not a prompt, and the payoff is brand-safe content and support you can actually audit.
TikTok Shop will outsell Target this year. Ninety-nine percent of sellers won’t feel it.
US TikTok Shop is forecast at $23.4 billion in 2026, bigger than Target’s or Costco’s e-commerce. But the top 1% of sellers drive about 60% of that, and half the headline stats are forecasts and second-hand claims. Read it with both eyes open.
Prompt engineering is dying as a job title and growing as a skill you cannot skip
The standalone “prompt engineer” role is fading, but the competency is now baked into marketing operations. What survived is not clever wording. It is structured prompting, real context, and prompts you actually test.
Let the AI write the script. Don’t let it decide what the script does.
Google Ads scripts still earn their keep in 2026 doing deterministic, boring work: budget alerts, zero-conversion flagging, out-of-stock pausing. AI can now draft them from a sentence. The judgment about the rules is still yours, and that’s the whole point.
Most of your retargeting is theater now. Here’s what still converts.
Third-party cookies stopped carrying your retargeting a while ago. The value moved to first-party audiences, server-side signals, and the automation layers inside Meta and Google. Here’s what to keep funding and what to quietly cut.
60/40 was an average. You turned it into a commandment.
The most cited number in marketing came from analyzing 996 campaigns and taking the mean. It was never a law, and applying it to a three-year-old brand burning runway is how good research becomes bad strategy. The rule is real. The dogma is the problem.
AI won’t triple your conversion rate. It’ll move it 10-30%, if you aim it right.
The average e-commerce conversion rate is somewhere between 1.4% and 2.96%, which tells you almost nothing. AI personalization is a real lever, but the honest number is a 10-30% relative lift, not the miracle in the case study. Here’s where it actually shows up.
Past pieces worth revisiting.

Connect Google Analytics to an AI assistant, and let it do the reporting.

Labeling your AI creative is the law now. That it costs you sales is a rumor.

Build a Google Ads agent that can’t wreck your account.

Don’t panic-bid on your own brand after the spam update.
“An audit built on borrowed percentages is a sales document. One built on your own arithmetic is a decision.
Pick a lane
Popular right now
- 01
Most audit findings are assertions. These six leave evidence.
Data & Research · 6 min - 02
ACP and UCP are converging. The merchant work underneath was never the argument.
E-Commerce · 8 min - 03
Google’s 7% and the independent 16% never measured the same advertisers.
PPC · 7 min - 04
The watermark counts the words Claude chose. It does not say who wrote the page.
SEO · 7 min - 05
ChatGPT Ads reached Poland with the personalization switched off.
PPC · 6 min
Editor’s picks
- 01
I shipped an AI iOS app solo in 30 days. Writing the code was the easy part.
AI & Data · 13 min - 02
ADAM: four pillars, five levels, and the one you’re quietly starving
AI & Data · 9 min - 03
Most audit findings are assertions. These six leave evidence.
Data & Research · 6 min - 04
Google’s 7% and the independent 16% never measured the same advertisers.
PPC · 7 min
One letter, once a week.
Sharp coverage of media, tech, and AI business, written for professionals who run digital, not for those who read about it. No filler, no growth-hacking newsletters about growth-hacking newsletters.

Writes about media, tech, and AI business for people who actually run digital. Former agency lead. Skeptic of frameworks that read better than they perform.
Also works as a digital consultant, helping companies with the day-to-day of their digital, from planning to shipping, over at salecki.digital.




















































