The Marketing edition
This is a real edition from 27 October 2026, at the Practical level, in plain language.
Agents visit your website, ChatGPT shows ads in 89 countries, and Google and Meta change their terms.
Your website gets a third kind of visitor: the AI agent
Until now, the internet had two kinds of visitors: people and bots. Now a third kind is arriving: personal AI agents. They do things online for someone, such as booking a flight or hotel. Apps like Dots from OpenAI and Muse from Meta have released them on a large scale. Muse was downloaded 5 million times in its first month. Agent traffic on the web tripled from August to September.
Many websites keep agents out. Yelp does not allow AI to make restaurant reservations. Amazon blocks Muse's agents and is suing the AI company Perplexity. A judge recently ruled in Perplexity's favor, but Amazon appealed. Amazon says agents threaten its advertising business, because it gets harder to separate ads seen by people from ads seen by bots.
Most agents visit web shops, so there is money to be made there. Shopify therefore already gives agents their own checkout lanes, which skip the pages meant for people.
The writer fears the web will get worse for people. He mentions worse websites, more and more proof that you are human, and different prices for different agents. You also can't know whether ads influence your agent's advice. Professor Ethan Mollick expects agents to flood customer service so much that people can no longer get through.
Agents can also really help people, for example with unfair medical bills or a cheaper car insurance. Whether that works depends on rules that are still being written. The question is whether companies may block agents or let them negotiate.
Agree with your web team what counts as a successful visit for you, now that a visit of a fraction of a second can also be a success. Write it down before you decide whether to block agents or give them their own route, because then you also know what you are measuring. That way you can explain later why a low number of seconds on a page is not automatically a bad number.
A successful visit can last a millisecond
Jasper Boot writes in an opinion piece that many customers let a bot do their searching. He noticed this himself with ChatGPT's voice mode. The bot visits your website on behalf of the customer, reads the information and gives advice. The customer then no longer visits your site. This distorts your statistics. An agent doesn't need four minutes. It reads the page very fast and leaves. Your report then shows a session duration of almost zero seconds and a high bounce rate, even though it may have found exactly what it was looking for. According to Boot, an agent has no use for mood videos, a nice button color or a beautiful brand story. It looks for overview and accessible data. Is your stock status in a slow, interactive module, or your price behind a required contact form? Then it gets stuck and the purchase goes to a competitor. His advice: look at your KPIs differently, win people over with a good story and help the machine with clear information. Read more →
ChatGPT ads now in 89 countries
OpenAI is expanding ChatGPT Ads by 26 countries, so the ads are now available in 89 countries. OpenAI started with ads a little over eight months ago. Since then, more than 100,000 advertisers have run campaigns. According to OpenAI, more than 1.2 billion people use ChatGPT every week. New countries include Argentina, Chile, Colombia, Peru, Nigeria, South Africa, Serbia, Ukraine, Cambodia, Bangladesh and Pakistan. You can take part through technology partners or yourself via Ads Manager. The ads appear only in the Free and Go plans. Plus, Pro and Enterprise stay ad-free. OpenAI says ads are clearly labeled. According to the company, conversations stay private from advertisers and customer data is not sold. As a next step, OpenAI wants to develop more formats, goals, buying options and measurement options. What that will look like has not been worked out yet. More information is at ads.openai.com. Read more →
How ChatGPT chooses which ad you see
OpenAI explains how ads in ChatGPT work. A normal ad system starts with a clear action, such as a search query. In ChatGPT, people come with a goal and sharpen it during the conversation. That is why OpenAI built its own system. A specially trained language model runs the whole process. It finds ads, ranks them, decides whether they are useful and runs the auction. It remembers information within a conversation and also across several conversations. Choosing ads runs separately from creating answers. This is meant to stop ads from influencing the answers. For product ads, there is a separate index that also takes positive and negative reviews into account. So good reviews can decide whether your product rises to the top. According to OpenAI, ChatGPT Ads reached annualized revenue of 1 billion dollars in less than 200 days. Later, OpenAI wants to try new formats, such as talking directly to businesses through an ad. Read more →
One month of ChatGPT ads at AFAS: high conversions, no proof yet
AFAS has been advertising in ChatGPT for a month. Conversions are higher than the company expected, although the writer gives no number. He thinks it is because of the moment at which you reach someone. In a conversation, a user has often already explained what is going on, so the ad fits a concrete need. The writer himself says this is not proof. One month is too short to judge the value of business leads. A completed form is only a first result. In the campaign settings you add hints. The more concrete the situation, the better. The landing page must be just as sharp. AFAS also manages the campaigns from the chat, with the ChatGPT Ads Manager plug-in. With it you can check ads, view performance and have changes made. Budget limits belong in the campaign settings. The next step is finding out which campaigns bring valuable inquiries. Read more →
Google weakens its promise about Gmail and ads
Google changed its privacy policy on 16 October. It used to say: 'We don't show you personalized ads based on your content from Drive, Gmail or Photos.' That firm promise is now worded more weakly. Android Authority was the first to spot the change. It concerns Personal Intelligence, which Google launched in early 2026. With it, you share data from several apps with Google's AI products, including AI Mode in search. AI search results can contain ads. The Gemini app has no ads at the moment. According to Ars Technica, shared data could therefore count in ad targeting. Google says ads are not based 'directly' on your inbox or photo library, that linking is voluntary and that it does not use sensitive information for ads. This is a policy change, not a new ad tool. So your campaigns don't change directly because of it, but clients may ask questions about it. Read more →
Meta removes the word 'train' from business terms
Meta has removed the word 'train' from its new commercial terms of use. They take effect on 30 October. Lawyers warned last week that Meta could use this to scrape, without limit, websites that are listed on business pages. Experts told Nieman Lab that Meta may in this way avoid millions in licensing costs for news organizations. Meta says the crawling terms apply only to ad and business products and that it does not change how it uses web crawlers. Content owners can state their wishes with robots.txt. Meta did not tell users that the text was changed. The experts think the changes don't go far enough. Lawyer Matt Topic says publishers should stay concerned. If you have a Meta business page or ad account, the terms apply to you too from 30 October. Read more →
Pennsylvania requires a label on ads made with AI
The US state of Pennsylvania has a new law on AI in advertising. Governor Josh Shapiro signed the Artificial Intelligence Deceptive Advertising Act on Thursday. An ad that is heavily edited or made by AI must get a label in the same form as the ad itself: photo, video, audio or text. It applies to big companies and to a restaurant's flyer. Where the ad was made does not matter, if it is also shown in Pennsylvania. The law leaves a lot of room. What 'heavily edited' means is not fixed. Parody, satire, commentary and criticism are exempt, as is anything 'clearly fictional'. A label seems to be needed only if the AI content gives a wrong picture of a product or service. What is 'wrong' and who decides is unclear. The law does not apply in the Netherlands or the rest of the EU. Read more →
Ask your web manager what is in your site's robots.txt file. Meta points content owners to this file to state their wishes, and then you at least know which wishes your site is currently passing on.
AI and tech, made simple.
This is a sample of the Marketing edition. Subscribers get it in their inbox every week.
Subscribe
inklaretaal · Amsterdam · © 2026
This is the generic Marketing edition. Subscribe and you get it made to measure: tuned to your level, your job and the topics that matter to you. That way every week gets a little more personal.
Start for free