AI developments: implications for marketing and business

AI is moving beyond a standalone chat into environments where you search, create content, analyse data and carry out work. The important question is not simply which model is new. It is what changes in how customers find information and how you organise your work.

This overview for 23 October 2026 brings the relevant developments together. Sources were checked through 21 October. An announced feature is not automatically available in your account, and a limited test is not a general launch.

The overview at a glance

The common thread is closer integration: models are moving into search engines, marketing tools and business data. Control becomes more important when AI can use systems rather than just write about them.

1. Gemini 4 Argon: more than a new model number

Google introduced Gemini 4 Argon on 13 October. Its announcement says the model is accessible globally through the Gemini API in AI Studio and the Gemini platform. In the Gemini app and Search AI Mode, access is for Google AI Pro and Ultra subscribers. That is not the same as free access for everyone.

Google emphasises complex multi-step tasks, coding and document analysis. Published performance claims and examples come from Google; they do not prove that your campaign or business process will see the same improvement.

For marketing and business: the interesting opportunity lies in tasks that combine information. Examples include comparing customer questions with product information, structuring a brief or preparing code for an internal tool. Supply clear sources and assessment criteria. Do not base exact calculations or consequential actions on an unchecked model answer.

The difference between the model and the application remains important. An LLM can process text and context; integrations and software determine which files or systems it can actually use.

2. ChatGPT: from an answer to an interactive tool

In its 7 October announcement, OpenAI describes Intelligent UI with GPT-6: responses can combine text, visual and interactive elements. Its availability information describes a rollout to eligible accounts. 'For everyone' is therefore not a careful summary of access.

For this development, we could check the public announcement and availability information only through indexed passages. We therefore do not reproduce an extensive list of subscriptions or countries, or claim that every described feature already works everywhere.

For marketing and business: the boundary between content and tools is narrowing. Rather than just explaining a comparison, an application may offer an interactive way to explore it.

This does not make useful tools on your own website redundant. A calculator, product comparison or enquiry process needs to match your actual offer, use correct data and lead to a next step. An attractive interface with unreliable content remains an attractive mistake.

3. Gemini brings brand style, marketing and business tasks together

On 20 October, Google announced a new business landing page in Gemini. The description covers brand style, marketing visuals, customer insights, data analysis and connections to existing applications. Google says the features are beginning to roll out globally to Gemini app users over the following days.

The announcement names Google Business Profile, Google Ads and Workspace, alongside external applications such as Canva and Shopify. Google also describes supplying a website, brand files, fonts or colours to make visuals more consistent. This does not mean every integration is already present in every account or has unrestricted data access.

For marketing and business: this may reduce the back-and-forth between briefs, files and execution. A useful trial is turning an existing campaign brief into draft variants using your brand guidelines as context. Check product representation, claims, spelling and consistency before publishing anything.

The strategic choice remains yours: what do you want to say, to whom and why? More variants are only useful when you know which differences you want to test. Read about content marketing to explore the connection between message and execution.

4. Google makes source links more visible in AI search

On 15 October, Google described changes to AI Mode and AI Overviews. These include more links placed higher in answers, more prominent related-article links and expansion of news carousels. A dedicated presentation for official websites is a test, not a guaranteed placement for every site.

For online marketing: AI search and your own website are not opposites. A search answer may partially resolve a question, while source links can still bring visitors to a relevant page. Google does not promise more traffic or better rankings for individual sites with this announcement.

Make clear what a page answers, what supports a claim and how current the information is. Product data, conditions, contact options and explanations need to be correct on your own site too. Writing only to secure an AI mention is not a substitute for a useful destination.

Look beyond visit counts: which pages are found, which visitors click through and which enquiries or purchases follow? Measuring AI visibility requires several signals; one search query is not a complete picture.

5. Visual ChatGPT ads: a test, not a general channel yet

On 5 October, OpenAI announced a visual advertising test around image generation. The checked passages describe a start later in October in the United States, with a limited group of advertisers and for Free and Go users. This does not confirm general availability in the Netherlands.

The description says advertisements will be labelled and separate from the generated image. OpenAI also mentions expanded measurement, but we could not independently read its full details. We therefore do not attach specific measurement capabilities to this test.

For online marketing: this matters because commercial visibility can appear during a task, not just a traditional search. It is something to monitor, not a reason to move your Dutch media budget based on this announcement.

A new channel ultimately needs to answer the same questions as an existing one: does it reach the right need, can you assess the result and does the destination meet expectations? Compare channels on those points rather than the amount of news about them.

6. AI agents need management, not just better prompts

Microsoft documents how Agent 365 associates risk signals with individual agents. Administrators can restrict access, block an agent or escalate a signal for investigation. Separate documentation describes approval before an agent becomes available across an organisation. This confirms the capabilities, not a separate new launch on 19 October.

For business: once an agent can read customer data, send messages or change settings, answer quality is no longer the only issue. You also need to know who is responsible, which actions are allowed and how to intervene.

A marketing agent preparing a report needs different permissions from an agent adjusting advertising budgets. Make that distinction explicit. Do not let publishing, sending and important changes quietly become part of a task that began with 'analyse this'.

Read more about AI agents and Agent Harness: an agent's surrounding environment is at least as relevant as its model name.

7. More model choice in business cloud environments

On 21 October, Google Cloud announced OpenAI models in its model offering. The important restriction appears in the announcement itself: rollout begins as a private preview for selected enterprise customers. It does not confirm immediate access for every Google Cloud customer.

Google describes purchasing through Cloud Marketplace and consolidated billing. Some additional security capabilities are labelled 'coming soon'; we do not treat them as already available.

For business: model choice can increasingly fit within existing procurement and management. This is particularly relevant if you already build applications and want to compare multiple models. By itself, it does not justify moving your systems.

Assess a model against your task: quality, speed, cost, data use and management. A smaller or cheaper model may suit a bounded task better than the latest model attracting the most attention.

What is the conclusion for your organisation?

The main change is not that you need more AI tools. AI is gaining access to the places where customers search and where work happens. This makes good content, clear data and control over execution more important.

Our interpretation of these developments for marketing and business is:

• Keep your website useful as both a source and a destination. Check the pages where customers decide or get in touch.
• Test one bounded task. Compare it with your current workflow and include review and correction time.
• Keep brand guidelines and source material in an identifiable place. A model cannot reliably fill missing context.
• Separate proposals from execution. Give an agent only the permissions its task requires.
• Check availability before planning around it. A test, private preview and global rollout are different things.

Large funding rounds say little about how useful a tool is for your organisation. The unconfirmed valuation in the inspiration material is therefore not central to this overview. Working features and their limitations matter more than a possible stock-market debut.

Want to turn these developments into a concrete question? In a strategy session, we can discuss what fits your marketing, data and way of working.

Sources and information cutoff

Information cutoff: 21 October 2026. Publication date for this overview: 23 October 2026. Access and rollout restrictions are stated alongside each development. Business applications in this article are our interpretation, not measured results or vendor guarantees.

• Google: introducing Gemini 4 Argon, 13 October
• OpenAI: GPT-6 and Intelligent UI, 7 October, and ChatGPT availability — checked through indexed passages; full pages could not be read
• Google: Gemini business tools, 20 October
• Google: source links in AI Mode and AI Overviews, 15 October
• OpenAI: advertising format and measurement, 5 October — checked through indexed passages; full page could not be read
• Microsoft: agent risk signals and agent approval
• Google Cloud: OpenAI models in private preview, 21 October