Welcome to issue #46 of the Marketing with AI Weekly Roundup. Published every Sunday, it shares the top stories, playbooks, data and quick hits from the world of Marketing with AI over the last 7 days.
Subscribe to receive the week's most important AI marketing stories direct to your inbox.
NEW: For 6 months I've asked every marketer I train the same question: where could AI save you time? Most aren't sure. So I turned everything those trainings taught me into a $5 tool that tells you. Your tasks, your hours back, your first workflow to build.
Last weeks edition was about disclosure, whether you have to tell people you used AI. This week the questions got more personal: what your numbers will say on 18 August, and what your job will look like in two years.
On 17 August, nine days from now, Google changes how Smart Bidding treats campaigns that are limited by budget. If you run a lead gen campaign that has quietly been beating its target for months, that campaign is the one that gets hurt, and nothing will land in your inbox on the day it happens.
On Tuesday the American Marketing Association published the clearest data anyone has produced on which marketing roles AI is actually eating, and its answer is not the one the industry keeps telling itself.
And on Wednesday the most widely followed independent LinkedIn researcher reported that the platform has started scoring individual comments, with AI-written ones counting against you.
Elsewhere, Adobe put more than seventy of its tools inside ChatGPT with a free tier that needs no Adobe account, three Meta Ads reports went dark without an error message, and the IAB published the first industry standard for measuring AI visibility, arriving a good while after the money did.
Let’s get into it…
🔥 Top Stories
1. Google resets what your bid targets mean on 17 August, and the campaigns beating their targets are the ones that get hurt
From 17 August 2026, Google Ads changes how Smart Bidding behaves on campaigns that are limited by budget. Today, a budget-capped Target CPA or Target ROAS campaign often overperforms, delivering acquisitions at a cost well below the number you typed in, because the budget cap is doing the constraining rather than the target. From 17 August, the system optimises toward the target you actually set, so those campaigns drift upward toward it.
Two things this is not. Campaigns with no budget ceiling are unaffected. And the auction itself is not changing, this is a bidding change only. Google has publicly pushed back on the reading that the update amounts to letting the system spend more money, stating it will not lead to spend increases.
In scope: Target CPA, Target ROAS, and Target CPC for Demand Gen, across Search, Shopping, Performance Max, Demand Gen and Travel campaigns, and across Google Ads, Search Ads 360, Display & Video 360, Google Ads Editor and the API. Not affected: App campaigns, Video reach and Video view campaigns, and any campaign not limited by budget.
The change was announced back on 15 June by Google Ads Product Liaison Ginny Marvin, and we have carried it in Looking Ahead since 11 July. What arrived this week is the operational guidance, a Q&A from Marvin answering the questions advertisers were actually asking, covered on 6 August. The deadline is now nine days away and the how-to finally exists.
Why it matters / what to do: picture a lead gen campaign permanently capped by budget, delivering leads at £22 against a £35 target you set eighteen months ago and have not looked at since. Google is about to start treating £35 as the goal rather than the ceiling. That is a 59% increase in your cost per lead, arriving with no email on the day. Here is the actual to-do list. First, filter your account for campaigns with the status Limited by budget running Target CPA or Target ROAS, which is your exposure list. Second, put a column for actual CPA or ROAS over the last 60 days, not last week, next to the stated target. Third, flag every campaign beating its target by a wide margin, because where the two numbers are already close the impact is minimal and you can leave it alone. Fourth, decide per flagged campaign: lower the target to match what it genuinely achieves, set a number that reflects what works commercially, or keep the target because it is still right and accept the drift. Fifth, write down what you expect to happen before 17 August, so you are not reconstructing the comparison afterwards from memory. Then leave it alone, because the rollout takes a few weeks with a calibration period and panic-fixing on 18 August will teach you nothing.
Three caveats. Google’s own Bid Target Adjustment Tool, live since 6 July, will suggest changes for you, but do not apply them blindly, because it has no idea what your business can actually afford per lead. Performance Planner forecasts are explicitly unreliable between 17 and 31 August while the tools recalibrate, so do not commit budget off those numbers. And Google’s own FAQ warns against reactive fixes such as data exclusions or bid limits applied purely because of this update.
2. The American Marketing Association put numbers on which marketing jobs AI is taking, and the profession is deprioritising exactly the skills that survive
On 5 August 2026 the American Marketing Association released its 2026 State of Marketing Careers Report, a mixed-methods study built on a survey of 1,412 marketing practitioners conducted in December 2025 and January 2026, supplemented with interviews, job-posting analysis and secondary research.
The share of marketing job postings mentioning AI nearly doubled across 2025, from 8% in January to 15% in December. Over the same period total marketing job postings fell by more than 8%, and US marketing employment remains roughly 27% below its March 2020 level. The roles that lost the most ground were social media coordinator, copywriter and SEO manager. Senior leadership held up considerably better than middle management and individual contributor roles.
The AMA’s own summary is the line worth carrying: “Companies are not hiring less judgment. They are hiring less execution.”
The report scores marketing tasks against Stanford’s Human Agency Scale, running from H1, where AI handles a task alone, to H5, where AI cannot function without continuous human involvement. The disciplines it places at highest disruption are email marketing, SEO and paid media, on the grounds that they are rules-based and repetitive. Routine reporting, data collection and campaign monitoring sit at the automatable end, alongside scheduling and social monitoring.
Then there is the finding that should genuinely bother you. Marketers rated critical thinking, communication, collaboration and adaptability as less important in 2026 than they did in 2025. Those are precisely the four capabilities the AMA’s own analysis places at the top of the human-involvement scale. The profession is actively demoting the things that are hardest to automate, at the exact moment the execution work is being automated out from under it.
Why it matters / what to do: this is the closest thing to an honest answer to “am I going to be fine” that anyone has published this year, and the answer is conditional rather than reassuring. Action: look at the report’s H3 zone, where humans and AI working together outperform either alone, which the AMA names as content making and data storytelling. That is where your next quarter of learning should go, rather than into another tool. And if your job title is one of the three that lost ground, the practical move this month is to attach yourself to a decision rather than to a deliverable, because the deliverable is the part being commoditised. Three caveats worth holding: the AMA is a professional body with a commercial interest in marketers continuing to invest in professional development, the job-posting analysis is US-based, and the skills-importance figures are self-reported perception rather than measured outcomes.
Read more: https://martech.org/here-are-the-marketing-skills-ai-is-making-more-valuable/
3. LinkedIn has started scoring individual comments, and AI-written ones now count against you
On 5 August, Richard van der Blom published his latest LinkedIn algorithm findings, the research series most practitioners in this space treat as the standard independent reference. Four things in it matter.
Reach and engagement fell again. July was down 18% in reach and 16% in engagement compared with June.
LinkedIn is now rating individual comments, not just posts. More than ten factors decide whether your comment gets distribution or quietly dies, and one of them is whether the comment was partly or fully written by AI, which is now a negative signal.
Send in DM has overtaken saves and reposts as the strongest distribution signal. Testing across more than 800 posts and 2,400 DM sends found that posts sent privately saw roughly a 60% uplift, and the effect pushes distribution outside your own network.
And “Celebrate an Occasion” posts are being boosted hard, up to 2.6 times the reach they got in March 2026. That looks like a deliberate rebalance back toward network signals after months of complaints that the interest-based feed had buried people’s actual connections.
Why it matters / what to do: LinkedIn is the single most important organic channel for most B2B marketers, and last week it gave every member a button to report a post as AI slop. This week we learn the same logic has reached the comment box, which is where most B2B teams do their AI-assisted volume work. Comments are the part of LinkedIn activity people most often automate, and it is now the part with an explicit penalty attached. Action: stop running AI-generated commenting of any kind this week, and if you use a tool that does it, turn it off. Then change what you optimise for. Asking someone to like your post is worth much less than writing something a reader would send to a colleague, so make the post useful enough to forward rather than easy to applaud.
One caveat, and it is important. This is independent practitioner research published by a consultant on LinkedIn, not an announcement from LinkedIn, so treat the figures as well-sourced observation rather than confirmed platform mechanics. Worth adding, because a lot of it went around this week: the wider claims circulating that LinkedIn has just changed its algorithm to favour expertise and consistency mostly trace back to agency blog posts with no LinkedIn attribution at all, and the underlying feed rebuild they describe actually shipped in March 2026.
4. Adobe put more than 70 tools inside ChatGPT, and the guest tier needs no Adobe account
On 6 August 2026 Adobe released a single Adobe plugin in ChatGPT, replacing the three separate connectors it shipped in December 2025 with one plugin reaching more than 70 tools across Adobe Express, Photoshop, Firefly, Premiere, Acrobat, Lightroom, Illustrator, InDesign and Adobe Stock. You describe the outcome in plain language and the plugin picks the tools. It is available globally from launch, on ChatGPT web and the desktop app, and works across ChatGPT, ChatGPT Work and Codex.
The setup path: open ChatGPT, go to Plugins in the left toolbar, search “Adobe”, install the one labelled “Adobe Design, combine, and edit”, then start a new chat and type @Adobe. You can continue as a guest. Signing in with an Adobe account unlocks Firefly image and video generation, access to your Creative Cloud files, and saving work between sessions.
The workflows Adobe itself names are unusually marketing-shaped: apply one consistent style across a whole batch of campaign photos and save them to Creative Cloud, turn images into product or packaging mock-ups, cut a long video into a highlight reel and reformat it for Shorts and Reels, customise a template’s copy and colours then export a print-ready PDF, or upload a dataset and get back a finished PDF such as an event badge set or a catalogue. It can also search your Creative Cloud library by loose descriptions like style or mood, which is probably the feature you would actually use daily.
Why it matters / what to do: on a small B2B team the constraint on shipping campaign assets is rarely the idea, it is the twenty minutes of resizing, recolouring and exporting that sits between the idea and the post. That work has just moved into the same window where the brief was written. Action: take one asset you resize every single week, a webinar promo or a conference banner, run it through @Adobe as a guest, and compare the output against what your designer or your Canva template produces. Decide on the evidence rather than on the demo. Three caveats: it is web and desktop only at launch with no mobile, how much you can generate depends on your ChatGPT plan and Adobe has not published the per-tier limits, and guest access saves nothing between sessions.
📋 The Playbook
1. Turn your team’s internal conversations into a daily LinkedIn content engine
Cody Schneider returned to Greg Isenberg’s channel on 5 August with a second system, and this one needs no engineering at all. He sells a platform and implementation services for exactly this kind of work, so treat him as an operator with a commercial interest rather than a neutral tutorial. The method survives that.
The rule the whole thing rests on: never ask a model to invent the content. His words: “If you go and you try to just have the agent like think about this, you’re like, ‘Write good LinkedIn content.’ It’s going to be the most mid thing, or you’re going to get flagged for AI slop by LinkedIn’s new feature that just released this morning.” That is a direct on-camera reference to the story we ran last Sunday, and it is the same conclusion from the other direction.
The workflow, step by step:
1. Generate source material on a cadence. His method is a weekly one-to-one call with each person on the team, deliberately unstructured: “just like tell me everything that like you’ve learned in the last week.” No agenda required.
2. Or mine the source material you already have. Sales call recordings, Gong transcripts, the sales channel in Slack, Notion, internal comms. His example of what is hiding in there: a prospect explaining why they did not buy, which he calls “an unbelievable piece of content.”
3. Extract the insights, then write one post per insight. He is candid that the writing model barely matters here, saying Claude Sonnet is “probably good enough on the writing side”, because the raw material is doing the work.
4. Publish across several named people, not just the company page, since personal accounts carry organic reach that brand pages do not.
5. Feed the performance data back into the loop, so the next writing cycle knows which insights earned attention. His two instruction words, used literally: snowball and remix.
6. Recycle the winners on a 90-day cycle. “If you look at my Twitter post as an example or even my LinkedIn, it is the exact same thing remixed every 90 days, like full stop. I can’t post it every day. You post it every 90 days.”
The number that makes this a budget conversation. He prices earned reach against paid: “on LinkedIn, it’s like $22 per thousand impressions is the average. Every post that you get, even with an account that’s like 500 followers, you can get a thousand impressions. That’s like $20 that you just like put into your pocket for free.” That is a defensible way to put a figure on organic in a planning meeting, though the $22 CPM is his number rather than a published LinkedIn rate, so attribute it to him.
The provocation, included because it deserves arguing with. “I actually think the social media manager job, like full stop, it’s already dead.” Isenberg pushes back and they land on evolved rather than dead. Set that against this week’s AMA finding that social media coordinator is one of the three roles losing the most ground, and it is less throwaway than it sounds.
And the alternative for anyone who will not build a personal brand: build a topic page instead of a personal one. Isenberg’s worked example is Julian Shapiro, who ran his agency Demand Curve behind an account called @GrowthTactics rather than a branded handle.
Do this now: record your next internal team call, pull three insights out of the transcript, and post them under three different named people this week. That is the entire system at its smallest possible scale, and it costs nothing.
Read more:
2. Set up Obsidian as the shared file layer that both Claude and ChatGPT can read and write
Paul J Lipsky published this on 5 August, and it is the most useful beginner build of the week because it solves a problem most people have not noticed they have: everything you teach an AI tool currently lives inside that tool. No sponsor is disclosed and the Obsidian link in his description is the plain product URL.
The argument in his words: “Obsidian operates entirely on local markdown or text files stored directly on your computer. This gives you complete ownership and control over your notes and you could switch to another tool at any time.” And on why it pairs so well: “tools like Claude Cowork and ChatGPT love markdown files. And since Obsidian uses markdown files too, it’s kind of a match made in heaven.”
The setup. Download from obsidian.md and create a vault, which is simply a folder. Save it somewhere backed up, he uses iCloud Drive and notes Google Drive or Dropbox work equally well. Formatting is plain markdown: # plus a space for a heading, - plus a space for a bullet, - space [ space ] space for a checklist item. Organise with folders and subfolders, right-clicking a folder to create a new note or subfolder inside it.
The thing that surprises people: the vault is an ordinary folder in Finder, so you can drag any file straight into it and it appears in Obsidian. PDFs and images render in place.
One settings fix you will need. Excel files do not show up by default. Go to Settings, then Files and links, and switch on Show all file types. They still will not open inside Obsidian, but at least you can see they exist. While you are there, turn on automatic updates under Settings, General.
Connecting Claude. With Cowork selected, use the dropdown, choose add a folder, navigate to your vault, select the whole vault (or a single subfolder if you want to limit what it can reach), open, then always allow. His demonstration prompt is worth copying: “create a new note for me in this vault that is an index of everything in the vault.” Claude writes a vault index file back into the folder, with working internal links that click through to the real notes.
Connecting ChatGPT Work. Select Work at the top, choose your project, new project, select your vault folder, open, then create project. His framing: “projects are just folders.”
The trick that makes it usable day to day. When you want the assistant to work on the specific note you are looking at, click the three dots at the top right of the note, hover copy path, click from vault folder, and paste that path into the chat. His example instruction after pasting: “Restructure this file with nicer headings, emojis, checklists, etc.”
Do this now: create a vault, drop in your positioning document, your ideal customer profile and your brand voice notes, point Claude at it and ask for an index. You have just built a knowledge base that both Claude and ChatGPT can read, that lives on your own machine, and that neither of them owns.
Read more:
🛠️ New Tools & Features
1. HubSpot published a head-to-head against Ahrefs, with real prices on both
On 6 August HubSpot published a direct comparison of HubSpot AEO against Ahrefs Brand Radar, the two mainstream tools for tracking whether AI answer engines mention your brand. HubSpot AEO is $50 a month, or $45 annually, covers ChatGPT, Gemini and Perplexity only, and is wired into the CRM so its prompt suggestions come from your actual customer data. There is a free 28-day trial with 25 prompts and no credit card. Ahrefs Brand Radar covers seven platforms, adding Google AI Overviews, AI Mode, Microsoft Copilot and Grok, and is built on a database of 406 million real monthly search prompts. Standalone it runs $199 to $699 a month, but it is bundled free into Ahrefs plans from $29 a month, so if you already pay for Ahrefs you effectively have it. Worth knowing: Ahrefs added Claude tracking to Brand Radar at the end of July, as a custom-prompt-only source that costs eight checks per update against one for other platforms. Neither product tracks everything, and the honest split is breadth versus action. Caveat, and it is the obvious one: this is HubSpot’s own comparison of its own product against a competitor, and while the Ahrefs pricing checks out against Ahrefs’ published tiers, the framing does not.
Read more: https://blog.hubspot.com/marketing/hubspot-vs-ahrefs-aeo
2. Microsoft Clarity now groups your AI citations by topic
Announced in Microsoft Advertising’s first monthly product newsletter on 4 August, Topic Insights inside Microsoft Clarity’s AI Visibility reporting groups your AI citations by subject rather than listing them individually, so you can see which topics AI systems associate with your brand and where the holes are. Three metrics come with it and the vocabulary is worth learning. Grounding queries are the retrieval searches an AI system runs before it writes an answer, which is not the same as what the user typed. Citation share is how often your domain appears as a source. Share of authority is how often you are cited compared with competing sources, which is a genuinely competitive number in a category where most tools only report your own presence. Clarity is free, which makes this the cheapest topic-level AI visibility data available without buying a subscription.
3. Perplexity put Statista, CB Insights and PitchBook inside its research agent
On 6 August Perplexity announced Premium Sources, bringing paywalled professional research directly into Perplexity Computer. At launch that means Statista, CB Insights and PitchBook, with CB Insights alone contributing more than 8,000 research reports. Market sizing, category trend data and competitor funding history are three things B2B marketers routinely approximate in a deck because the real sources cost five figures, so a Statista or PitchBook citation in a positioning document is a different quality of evidence. The pricing genuinely does not reconcile and you should check your own plan before relying on it. Perplexity’s enterprise guide describes Premium Sources as built into your subscription at no extra cost, while a separate Perplexity post describes a much tighter allowance of three premium searches a month on free, five on Pro and ten on Enterprise Pro, and names a different set of providers. Those two descriptions cannot both be right.
Read more: https://www.perplexity.ai/hub/blog/announcing-premium-sources
4. Gemini Notebook’s cloud computer finished rolling out, and Workspace can now keep it fed
Two connected updates. On 4 August the upgraded Gemini Notebook experience was confirmed as fully rolled out to all Google AI Pro subscribers on the web, nineteen days after it was announced. That means every notebook now has a secure cloud computer that can write and run real code against your own uploaded sources, returning charts, PDFs, spreadsheets and slide decks rather than descriptions of them. The 20-second test for whether your account has it: upload a table and ask for a chart or a spreadsheet, and if you get a downloadable file it is active. Then on 6 August Google Workspace added an “Add a source to Gemini Notebook” step to Workspace Studio, so an automation can keep a notebook current on a schedule by feeding it Drive files, YouTube URLs or web pages. The second half is the more interesting one for marketers: a competitor-monitoring notebook that refills itself rather than going stale after a fortnight. Caveat: Workspace Studio steps are a Workspace feature, so that half does not apply to personal Gmail accounts.
Read more: https://workspaceupdates.googleblog.com/search/label/Gemini%20Notebook
5. Microsoft’s Ad Preview Hub reached Performance Max, with Bing search previews
Also in Microsoft Advertising’s 4 August product newsletter: the Ad Preview Hub now supports Performance Max campaigns, and adds a Bing search results preview alongside the existing MSN and Outlook placements. You get a shareable public link that expires after 30 days. This is the least glamorous item here and it is on the list anyway, because seeing what your ad actually looks like across placements before it goes live is a real marketing job that Performance Max has made unusually hard. The shareable link is the useful part for anyone who has to get creative signed off by someone without a Microsoft Ads login.
Read more: https://ppc.land/microsoft-performance-max-campaigns-gain-ad-previews-across-3-publishers/
📊 Research and Data
1. Forrester: 88% of B2B marketing organisations have adopted AI, and the foundations have not kept up
Published 5 August and drawn from Forrester’s Marketing Survey 2026, covering more than 1,000 B2B marketing decision makers, this is the largest and most credible B2B sample of the week. 88% of B2B marketing organisations have adopted AI tools or built their own. Alongside that, 85% say they run integrated campaigns and 90% of those say integrated programmes deliver more value than separate initiatives. Yet the same leaders report unclear AI strategy, difficulty measuring impact, data infrastructure problems and uncertainty about where AI should even be applied. Forrester’s framing: “marketing’s ambitions, priorities, and accountability are expanding faster than the underlying capabilities required to support them.” The named top challenges are poor data quality and accessibility, difficulty measuring performance, and insufficient insight for decision-making.
What to do: adoption has stopped being a differentiator, because almost everyone has adopted. The recommendation that follows is unglamorous and correct: treat data infrastructure, credible measurement and clear ownership as strategic investments rather than housekeeping. Practically, if you are choosing between buying another AI tool and spending a month making one dataset clean and reachable, the second is now the competitive move. Caveat: Forrester sells advisory services on precisely this problem, and the full methodology sits behind the report.
Read more: https://www.forrester.com/blogs/b2b-marketing-is-moving-faster-than-its-foundations-can-handle/
2. TransUnion: 89% will spend more on AI, 36% think their data can take it
Also published 5 August. TransUnion commissioned United Talent Agency’s brand advisory division to survey 100 senior marketing and technology leaders, all director level or above, all at US enterprise brands with marketing budgets of at least $50 million, and all already using AI in core marketing workflows rather than just for productivity. TransUnion calls the result the AI Confidence-Readiness Paradox.
89% expect to increase spending on AI-enabled marketing over the next 12 to 24 months and 64% are confident of hitting their AI goals. But only 42% rate their people readiness as high and only 36% say the same of their data and processes. Just 13% have fully scaled AI across the enterprise, and 53% report meaningful ROI. 65% measure AI’s impact primarily through estimated cost or time savings.
The split that tells the story: 75% say AI has reduced manual effort or time spent, but only 44% have reduced media waste, 35% have improved conversions or response rates, and 30% are reaching more of their target audience at similar spend.
What to do: AI is reliably saving time and unreliably making money, and two thirds of teams are measuring the thing it is good at rather than the thing they are paid for. Action: pick one AI-assisted activity and measure it on a business outcome instead of hours saved. If you find you cannot, that is itself the finding, and it is a more useful one to take to your leadership than another efficiency number. Three caveats, all worth stating: the sample is small at 100, it is enterprise-only with $50m-plus budgets so it does not describe a ten-person B2B team, and TransUnion sells identity and marketing data, which makes “marketers have a data problem” a conveniently placed conclusion.
3. Marketers are now putting 24% of search and content budgets into AI visibility
Reported by Digiday on 7 August, pulling together several vendor surveys. From Fractl: marketers route roughly 24% of search and content budgets to AI visibility work, 82% have allocated at least something, 18% allocate nothing at all, and 43% put more than a fifth of that budget into it. By function, performance marketing at 32% and SEO at 31% spend the highest share, while brand at 21% and content at 20% spend the least.
Two other numbers in the same round-up are worth pairing with it. 10Fold found 52% of B2B marketers now rate AI answer engines as their most effective distribution channel, against 29% for organic search, while 41% have only a quarter to a half of their content optimised for AI discovery. And Similarweb data put AI search at 35% of initial product discovery against 13.6% for traditional search, with the gap narrowing at the purchase stage to 24.3% against 22.1%.
What to do: this is the “am I behind” number for the week, and the 10Fold split is the sharper half. B2B marketers are already rating answer engines above organic search as a distribution channel while most of their content is not built for it, which is a gap you can close with editing rather than budget. Caveats: Fractl and 10Fold are both agencies with a commercial interest in AI visibility work, neither sample size nor fieldwork dates were published in the coverage, and the Similarweb figure is a behavioural estimate rather than a survey. Attribute all three to the Digiday round-up and be explicit that the samples are unstated.
💡 Quick Hits
1. The IAB published the first industry standard for measuring AI visibility.
Released 3 August, “Measuring Visibility in the AI Era” is a 36-page framework built on four levels, Presence, Prominence, Portrayal and Persuasion, and it splits measurement into directional and decision-grade tiers across nine criteria, treating anything under 50 queries per programme as exploratory. It states plainly that directional data is not sufficient for budget allocation, and notes that more than 20 companies now sell AI visibility tools on methodologies that can return different answers for the same brand. Its own headline number: only 16% of brands systematically track AI visibility today.
Read more: https://www.iab.com/news/iab-releases-measuring-visibility-in-the-ai-era/
2. Google will auto-upgrade some Search campaigns to AI Max on 1 September.
An email went to advertisers on 5 August confirming that campaigns using automatically created assets or the campaign-level broad match setting will be converted. The defaults differ by trigger: automatically created assets brings search term matching and text customisation on, while campaign-level broad match brings search term matching only. To avoid it, switch the legacy setting off or enable AI Max yourself before then. Dynamic Search Ads are not part of this wave, that migration was pushed to February 2027.
Read more: https://searchengineland.com/google-to-auto-upgrade-some-search-campaigns-to-ai-max-484428
3. Three Meta Ads reports went dark on 6 August without an error message.
The device, hourly and frequency breakdowns became opt-in per ad account with Marketing API v26.0. Requests on accounts that never opted in return HTTP 200 with an empty data array, so no tool alerts you and your dashboard simply renders zeroes. The toggle sits under Additional Breakdowns in Ads Manager’s reporting view, there is no Meta help centre article for it, and both Supermetrics and Improvado issued breaking-change notices to customers. Most sources say the missing data does not backfill, so the sooner you switch it on the less you lose.
Read more: https://admakeai.com/blog/meta-ads-updates-august-2026
4. ChatGPT Atlas stops working tomorrow, 9 August.
Bookmarks, open tabs and history do not transfer automatically. Cookies and passwords can be exported to the ChatGPT desktop app and bookmarks to Chrome. It launched in October 2025 and does not reach its first birthday.
Read more: https://help.openai.com/en/articles/6825453-chatgpt-release-notes
5. Google search rankings moved sharply between 1 and 3 August,
…Picked up by third-party trackers and heavily discussed in the SEO community. As of 5 August Google had not confirmed an update and its Search Status Dashboard listed no incident. That is the fourth unconfirmed tremor since the confirmed spam update in late June.
Read more: https://www.seroundtable.com/google-search-ranking-volatility-august-1-41811.html
6. Search Console’s new generative AI performance data is more misleading than it looks.
Roughly eight weeks after the reports launched, the core problem is that the central-tendency formula treats position one inside an AI Overview as equivalent to position one in classic organic results, which distorts any trend you try to read from it. Presence metrics also arrive with no click or query detail attached.
7. A correction to something we ran on 25 July.
When we covered Claude’s Record a skill feature we flagged that Anthropic had published nothing about what happens to your recordings. That gap has now been filled by hands-on documentation rather than by Anthropic: Claude discards the raw audio and video after processing and keeps only the screenshots it needs for context, which live inside the Cowork task and disappear with it. Two details we should have carried at the time: recordings cap at about 10 minutes, and the feature is Mac only, with no Windows support and no availability on Free or Enterprise.
Read more: https://www.unite.ai/de/claude-cowork-review-record-a-skill/
8. HubSpot’s Agent Hub credit pricing has firmed up…
…closing a question we left open on 2 August. Published rates are $0.50 per resolved conversation for the Customer Agent, $1.00 per recommended lead for the Prospecting Agent, $0.10 per answer for the Data Agent, and around $10 per piece for the Content Agent. Included allowances run 500 credits a month on Starter, 3,000 on Professional and 5,000 on Enterprise, pooled account-wide with no rollover. Agent Hub itself remains free for Professional and Enterprise and is still in public beta with no general availability date.
Read more: https://www.hubspot.com/products/artificial-intelligence
9. Anthropic has added noindex and nofollow to shared Claude conversations…
which resolves the problem we reported on 2 August when shared chats began appearing in Google Search. The original cause was a robots.txt block sitting in front of a noindex header, so Google never crawled the pages to see the instruction.
10. Claude Code’s auto mode becomes the default permission setting on 14 August…
…for Pro, Max and Team users, unless a user or admin has pinned something else. It replaces repeated approval prompts with a classifier that checks each action for anything irreversible or destructive, falling back to manual approval after repeated blocks. Anthropic says it will not charge for the extra tokens the classifier uses.
👀 On Our Radar
Time is selling ads that only AI crawlers can see, and every bot read is billed as an impression.
Digiday broke the business story on 30 July, reporting that Time had partnered with adtech vendor Mobian on a format they call agent ads. On 5 August developer Vincent Schmalbach published the technical teardown that proved the mechanic, and The Register independently replicated it. Request a Time page as Chrome, Safari or Googlebot and you get the normal HTML, roughly 303,000 bytes. Request the identical URL in the same second as ClaudeBot, PerplexityBot or OAI-SearchBot and you get a stripped 13,000-byte markdown version with sponsored content baked in that no human ever sees. What sits in the machine-only version is a full Ally Bank FAQ, labelled sponsored, written in Ally’s own marketing language, with structured data and campaign-tagged tracking links. Ally Bank and the Project Management Institute are the first two advertisers.
The mechanic is the interesting part. Response headers carry a fresh unique ID on every single request paired with a no-store cache instruction, so every bot read is logged as a distinct ad impression, and a second header names the unit being counted: tokens fed into a model. Not a person, not a pageview. Mobian’s chief executive put the pitch plainly, that the aim is to influence the assistant rather than the reader, because with a human you only influence one person.
Ignore the buying side, which is not available to you. The useful part is free and takes ten minutes. Fetch one of your own key pages six times, changing only the user agent, as a normal browser and then as Googlebot, ClaudeBot, OAI-SearchBot, PerplexityBot and GPTBot, and compare the six responses. If your CDN, your CMS or a plugin is already serving assistants something different from what it serves buyers, you want to find that out yourself rather than when an assistant summarises the wrong version to a prospect. Two caveats: this is one documented example, and Time deliberately excludes Googlebot, which reads more like avoiding a cloaking penalty than a principled position.
Read more: https://www.vincentschmalbach.com/time-serves-ai-bots-a-different-website/
First, the money moved into AI visibility before anyone agreed how to measure it. This week produced the cleanest piece of arithmetic we have run in months. 82% of marketers now put budget into AI visibility, at an average of 24% of search and content spend. 16% of brands systematically track it. Those two numbers describe the same people. On the same day, the IAB published the first standard for what good measurement even looks like and drew a line most vendor dashboards do not cross, saying in plain language that directional data is not good enough to allocate budget against. Microsoft then shipped topic-level citation reporting for free, and HubSpot and Ahrefs went head to head on price. The industry is closing the gap, but it is closing it in the right order only by accident, and the practical consequence for you is that you can now hold a vendor to a document instead of a demo.
Second, the honest measure of an AI programme is what you count, not what you adopted. Forrester found 88% adoption across more than a thousand B2B marketing decision makers. TransUnion found 13% fully scaled and 36% who trust their own data. Underneath both sits the number that should sting: 65% of marketing leaders measure AI’s impact primarily by hours saved, while 75% report saving time and only 35% report better conversions. This is the third consecutive week this newsletter has landed on some version of “everyone adopted, almost nobody rewired”, and it is worth saying that out loud rather than dressing it up as a new insight. What has changed is that the framing can now be sharper. Adoption is finished as a story. The thing separating teams is whether they are measuring AI on the outcome marketing is actually paid for, and most are not, because hours saved is the easier number to produce.
Third, the work being automated and the work being valued are pulling apart, and marketers are backing the wrong one. The AMA put numbers on it: execution roles losing ground, judgment roles holding, and the profession itself rating critical thinking, communication, collaboration and adaptability as less important than it did a year ago. Read that against LinkedIn penalising AI-written comments, and against a week in which Adobe moved seventy production tools into a chat window. The production layer of marketing is being commoditised in public, quickly, and the response from a lot of teams has been to get better at production. The uncomfortable implication is the same one the research keeps arriving at from different directions: the highest-return thing available to most marketers right now is not a tool at all. It is being the person who decides what should exist and why.
🔮 The Big Picture
First, the money moved into AI visibility before anyone agreed how to measure it.
This week produced the cleanest piece of arithmetic we have run in months. 82% of marketers now put budget into AI visibility, at an average of 24% of search and content spend. 16% of brands systematically track it. Those two numbers describe the same people. On the same day, the IAB published the first standard for what good measurement even looks like and drew a line most vendor dashboards do not cross, saying in plain language that directional data is not good enough to allocate budget against. Microsoft then shipped topic-level citation reporting for free, and HubSpot and Ahrefs went head to head on price. The industry is closing the gap, but it is closing it in the right order only by accident, and the practical consequence for you is that you can now hold a vendor to a document instead of a demo.
Second, the honest measure of an AI programme is what you count, not what you adopted.
Forrester found 88% adoption across more than a thousand B2B marketing decision makers. TransUnion found 13% fully scaled and 36% who trust their own data. Underneath both sits the number that should sting: 65% of marketing leaders measure AI’s impact primarily by hours saved, while 75% report saving time and only 35% report better conversions. This is the third consecutive week this newsletter has landed on some version of “everyone adopted, almost nobody rewired”, and it is worth saying that out loud rather than dressing it up as a new insight. What has changed is that the framing can now be sharper. Adoption is finished as a story. The thing separating teams is whether they are measuring AI on the outcome marketing is actually paid for, and most are not, because hours saved is the easier number to produce.
Third, the work being automated and the work being valued are pulling apart, and marketers are backing the wrong one.
The AMA put numbers on it: execution roles losing ground, judgment roles holding, and the profession itself rating critical thinking, communication, collaboration and adaptability as less important than it did a year ago. Read that against LinkedIn penalising AI-written comments, and against a week in which Adobe moved seventy production tools into a chat window. The production layer of marketing is being commoditised in public, quickly, and the response from a lot of teams has been to get better at production. The uncomfortable implication is the same one the research keeps arriving at from different directions: the highest-return thing available to most marketers right now is not a tool at all. It is being the person who decides what should exist and why.
Tell me: What’s your biggest takeaway from this weeks news from the world of AI Marketing? Love to hear your thougths in the comments.
Thanks for reading the Marketing with AI Weekly Roundup! If you enjoyed this read, the best compliment I could receive would be if you shared it.