Manual Focus

James and Hannah talk through the shared AI training opt-out that Google and Microsoft said on 30 October 2026 they will honour, a robots.txt token defined in an IETF draft by Anthropic's Kevin Kelley. They cover how it works, why Microsoft now separates training from grounding, the traps that could cost a brand visibility in AI answers, and why most brands should allow training by default and opt out only on the paths that carry content they sell or license.Read the article and get the prompts: What the shared AI training opt-out means for brand websites

James and Hannah go through the first independent study of what ChatGPT ads return. Maximilian Kaiser and Christian Schulze's working paper, dated October 2026, finds that in US traffic a paid ChatGPT visit earned $0.89 against $2.45 for a paid Google Search visit, which puts a ChatGPT click's parity price at about $0.26. They cover how the study used OpenAI's country-by-country rollout, why the largest advertiser makes the result look better rather than worse, what it means for bids and budgets, why organic ChatGPT referrals still matter, and three jobs to hand to AI before Black Friday.Read the article and get the prompts: What the first ChatGPT ads returns study means for paid teams

James and Hannah go inside the messaging house for WattPlan, the cycling training app Manual Focus builds. It has six levels, from the two-word tagline Train. Adapt. down to the features, each naming the code behind it, and writing it that way caught five stale claims before they reached a store listing. They walk each level with its definition, the four readings of the tagline, the message path and routing table, the understood and consistent test, the three banned lists, the foundation, the App Store and Google Play limits, and three jobs to hand to AI if you are building a house for your own product.Read the article and get the prompts: The WattPlan messaging house, how we built it and what it does

James and Hannah go through Reddit's third-quarter letter to shareholders, published on 29 October 2026. Reddit plans a logged-in-only site and will start testing it this quarter in markets outside the US, after its search referrals fell by more than 1 million a week in October. They explain why Google is better placed than other AI engines to keep showing Reddit content. Brand teams lose a free route into search while paid teams may gain sharper targeting. The episode ends on what Reddit hasn't said, after three jobs to hand to AI.Read the article and get the prompts: Reddit plans a logged-in-only site, what it means for marketers

James and Hannah talk through Google's 28 October 2026 announcement that ads in AI Mode are scaling to ten countries now, with the UK, Brazil and Mexico due in the fourth quarter, and that multiturn ads have launched in the US. Google's help page says existing Search, Shopping and Performance Max campaigns are eligible automatically, there's no separate opt-out, and results are reported only in aggregate. They cover why the UK launch lands in peak trading, what it means for Shopping and search teams, and three jobs to hand to AI before December.Read the article and get the prompts: What ads in Google's AI Mode mean for UK paid search teams

The second long-form Manual Focus Basics episode, and the second half of the story that began with What is marketing. James and Hannah take a finished plan to real buyers, following a hypothetical UK software start-up that sells project-profitability software to architecture practices. They separate growth strategy from growth tactics, then work through eight steps, the ideal customer profile, a wide list from Apollo, AI research and scoring with Claude, the shortlist, messages from the communication hierarchy, landing pages built with Claude Code in GitHub, sequences, and measurement, before pulling it together in SOSTAC. They cover the UK rules on B2B email, Google's sender guidelines, GitHub's terms for hosting and the mistakes that most often break the method. The start-up is an illustration, not a real company.Read the article and get the prompts: Go-to-market automation, a basics guide with Claude and Apollo

A long-form Manual Focus Basics episode that pulls the series together into one answer to "what is marketing?". James and Hannah start from the Chartered Institute of Marketing's definition, then follow a hypothetical new cycling helmet from the first question, is there a market, through market research, the product requirements brief, the communication hierarchy and the benefits ladder, then into the channels, the launch campaign, the creative and the PR, and finally a SOSTAC plan to run and measure it. They finish on where AI fits at each step and the mistakes that most often break the chain. The helmet and its findings are an illustration, not a real launch.Read the article and get the prompts: What is marketing, a basics guide from market fit to SOSTAC

James and Hannah talk through Google's new help page on AI review summaries in Search ads, first reported on 24 October 2026. Google's AI now writes positive-only summaries of online reviews into Search ads in the US, UK, Australia and India, advertisers can't edit them, and the page describes no opt-out. They cover how the feature grew out of a year of tests, why rule 3.46 of the UK CAP Code raises a question for positive-only summaries, what it does to brand and paid teams, and three jobs to hand to AI this week.Read the article and get the prompts: What Google's AI review summaries in ads mean for UK brands

James and Hannah update last week's episode on Meta's website-crawling terms. Meta told Reuters on 23 October 2026 that the new Commercial Terms and Self-Serve Ad Terms will take effect on 20 November instead of 30 October, though the UK English preview still showed 30 October on 25 October. Meta has removed the word train, so the UK and US editions now match, but the licence to crawl and use test credentials on sites linked to your ads is unchanged and the terms still don't mention robots.txt. They cover why the robots.txt plan stands, who gains most from the extra weeks, and three jobs to hand to AI.Read the article and get the prompts: Meta's crawling terms delay, what UK advertisers should do now

James and Hannah go through the SISTRIX study published on 22 October 2026, which compared 96,418 pages to see which GEO tactics go with citations in ChatGPT, Google AI Mode and AI Overviews. Tables of contents, Article schema, real tables and author boxes went with more citations, llms.txt showed no measurable effect, and product pages did poorly. They cover how to read the title-year finding, what it means for online shops, why small GEO studies mislead, and three jobs to hand to AI.Read the article and get the prompts: What SISTRIX's 96,418-URL citation study means for GEO work