AI in Marketing Takeaways for Leaders: July 2026
July’s AI developments point to further change in the landscape.
AI is increasingly becoming the intermediary between people and information, customers and products, employees and business systems, and brands and the platforms representing them.
Customers are asking AI systems what to buy, where to go and which businesses to trust. Marketing platforms are using AI to decide what people see. Service agents are resolving enquiries and directing users towards their next action. Workplace agents are being given access to files, tools and business systems.
This creates significant opportunity. AI can make discovery more relevant, customer journeys easier and organisations more productive. But the more active AI becomes, the more control businesses may be handing over.
For leaders, the question is moving beyond what AI can produce:
Where should AI be allowed to influence or make decisions?
What should it be able to do on behalf of the business?
How can leaders see what is happening?
And where must human judgement, accountability and control remain?
Here are the AI in marketing takeaways leaders should be aware of this month.
1. AI discovery is fragmenting, and measurement is starting to catch up
AI-led discovery is becoming an established part of how people find information, explore options and evaluate businesses.
Similarweb’s analysis of June traffic suggests ChatGPT is now behaving more like a mature global internet platform than a fast-growing start-up. Its growth is becoming steadier because of its existing scale, while challengers including Gemini and Claude continue to grow rapidly. Gemini recorded around 2.86 billion web visits during June despite experiencing its first monthly decline since December 2024.
This does not mean Google Search is suddenly disappearing. It means customer discovery is becoming more fragmented.
The same person may now use Google, ChatGPT, Gemini, social platforms, marketplaces, review sites and a company’s website during one decision journey. Each environment may interpret the same brand differently.
Measurement is beginning to improve.
Semrush expanded its AI Visibility Index from an initial 2,500 prompts to an analysis of 126 million prompts across ChatGPT, Gemini, Google AI Mode and Google AI Overviews. Its research found that 45% of marketing leaders cannot accurately measure their brand’s visibility in AI-generated answers, while only 9% have tools covering all the relevant metrics.
The research also distinguishes between being mentioned and being cited. A brand may appear in an AI answer without its website being used as the supporting source. Equally, a business may provide information that AI platforms cite without the brand itself being recommended.
Google has also begun rolling out dedicated generative-AI reporting in Search Console. The reports show impressions, visible pages, countries, devices and performance over time across AI Overviews, AI Mode and generative features in Discover. They are initially available to a subset of websites and do not yet provide a complete view of clicks, queries or commercial outcomes.
This is progress, but it does not make AI visibility simple. Different platforms use different sources, produce different answers and may respond differently to slight changes in how a question is asked. Traditional search performance remains important, but it is no longer a complete measure of digital visibility.
Leadership takeaway:
AI visibility should become part of search, brand and customer insight reporting. Start with the important questions customers ask at each stage of their journey. Monitor:
Whether the brand appears
How it is described
Which competitors are recommended
Which sources are cited
Whether information is accurate and current
How results differ by platform
Whether AI visibility leads to useful traffic, enquiries or sales
The objective is not simply to appear more frequently. It is to be accurately understood, credibly supported and relevant when customers are making decisions.
2. Search, shopping and advertising are converging around conversational intent
Product discovery, shopping and advertising are increasingly being built into the same AI-led customer journey.
Google is introducing AI performance insights in Merchant Center to help retailers understand how products are being discovered across AI Mode, AI Overviews and Gemini. These insights are intended to show how product data influences visibility within conversational shopping experiences.
Google is also continuing to extend the reach of its shopping ecosystem. From July, Shopping ads and free listings can target a further 14 international markets. While that expansion is not itself an AI feature, it sits within a wider direction in which Merchant Center data, conversational discovery, product recommendations and eventual checkout are becoming more closely connected.
At the same time, OpenAI has shared some initial evidence from its ChatGPT advertising pilot.
The company says the rate at which users dismiss advertisements has fallen by 50% since the pilot began in February. OpenAI is treating dismissals as an early indicator of relevance, suggesting that ads aligned with the context of a conversation may feel more useful than traditional interruption-based formats.
It is still very early. Lower dismissal rates do not tell us whether ChatGPT ads will deliver competitive acquisition costs, profitable sales or meaningful scale. There is not yet enough evidence to treat conversational advertising as a replacement for established search, social or retail media channels.
But the commercial direction is increasingly clear. AI platforms are bringing together:
Customer questions
Product recommendations
Advertising
Product feeds
Pricing and availability
Reviews and supporting information
Checkout and transaction capability
This increases the importance of structured commercial information.
Product names, descriptions, categories, imagery, specifications, pricing, availability, delivery information, returns policies, reviews and loyalty benefits are becoming part of how AI systems decide what to surface.
The same principle applies to services. Clear propositions, audience information, pricing context, locations, availability, credentials, case studies and customer proof all help AI systems understand when a business is relevant.
Leadership takeaway:
Product and service data should be treated as a marketing asset, not simply an operational feed.
Retailers should review the completeness, accuracy and consistency of Merchant Center and other product feeds. Service businesses should make sure AI systems can clearly understand what they offer, who it is for and why the business is credible.
Conversational advertising is worth controlled experimentation, particularly where customer intent is high. But budgets should follow evidence. Measure commercial outcomes rather than platform engagement signals alone.
3. Customer discovery is becoming visual, conversational and continuous
The traditional search box is evolving. Google has redesigned Google Images around a more browsable and personalised experience. Its new home includes a dynamic gallery of images tailored to users’ interests and saved collections, making it feel closer to a discovery platform than a traditional list of search results.
Google is also introducing image generation within AI Overviews, improving multi-object recognition and allowing users to upload several images before asking detailed questions about them. This means people can increasingly search by showing rather than describing.
A customer might upload a room, an outfit, a destination, a product or a screenshot and ask:
Where can I buy something similar?
Which items are shown here?
What would work with this?
Which destination has this style?
What is the difference between these options?
Spotify’s new conversational experience points in the same direction. Eligible Premium users can type or speak to Spotify, refine their request through an ongoing conversation and take actions such as saving a song, following an artist or changing what is playing.
The broader change is not limited to images or music. Customer interfaces are moving from precise keywords, menus and filters towards natural language, images, voice and ongoing conversation. Instead of learning how a website or app is organised, the customer describes the outcome they want. The platform then interprets that intent and decides what to show or do next.
Leadership takeaway:
Businesses should review whether their digital assets are understandable beyond traditional text search. That includes:
Image quality and consistency
Product and service context
Image metadata
Video and audio content
Structured product information
Frequently asked questions
Comparison information
Clear calls to action
App and website actions that an AI assistant could trigger
Customer journey planning should increasingly begin with what people are trying to achieve, ask or show, rather than only which page, keyword or menu they might use.
4. Creative scale is making consent, provenance and disclosure operational issues
AI is allowing marketing teams to create, adapt and personalise content at far greater speed. But July also provided a clear warning that creative capability can move faster than customer consent and brand governance.
Meta launched a feature within its Muse Image tool that allowed public Instagram content to be referenced when creating AI-generated images. The feature was withdrawn within days following criticism over privacy, automatic inclusion and the potential creation of non-consensual digital replicas. Meta acknowledged that the feature had missed the mark.
The incident matters beyond Meta. Platforms increasingly have access to large quantities of brand, employee, customer, creator and public content. The fact that a platform can technically use an asset does not mean the people represented in it have meaningfully consented to every possible use.
Google is moving in a different but related direction by increasing disclosure. Its new advertising transparency features allow users to see when an advertisement has been created or materially edited using AI. Ads produced using Google’s own tools can be labelled automatically, while advertisers using external AI tools are responsible for providing the appropriate information.
As AI production expands, marketing teams will need to answer practical questions:
Which assets may be used for AI generation?
Do agreements with employees, customers, creators and partners permit that use?
Can a person’s face, voice or style be reproduced?
When should AI-generated content be disclosed?
Which platform automation settings are enabled by default?
Who reviews generated variations before they are published?
How can the source and approval history of an asset be demonstrated?
This is not only a legal or compliance issue. Poor consent practices can damage trust. Inconsistent disclosure can create suspicion. Automated creative can change products, claims or context in ways that weaken the brand.
Leadership takeaway:
Consent, provenance and disclosure should be built into the creative workflow.
Create approved asset libraries, record usage rights, agree when disclosure is required and review the AI settings enabled across advertising and social platforms.
AI should make responsible production faster. It should not make approval, ownership or accountability less clear.
5. AI agents are moving from assistance to action, while controls lag behind
OpenAI’s GPT-5.6 launch shows how quickly AI is becoming more capable of completing complex, multi-stage work.
The new model family is designed to coordinate tools, process intermediate results and run concurrent sub-agents. ChatGPT Work is also positioned as a workplace environment that can reason across reference files, templates and connected business context to produce finished materials and complete longer tasks.
The opportunity is significant. Agents could prepare reports, update CRM records, analyse performance, create campaign assets, reconcile information, conduct research and coordinate tasks across several systems. But greater autonomy increases the consequences of a mistake.
Following the launch of GPT-5.6 Sol, several users publicly reported cases in which the model deleted files, databases or other resources without the approval they expected. These individual reports do not establish how widespread the problem is, but OpenAI’s own system documentation had already identified the possibility of the model becoming over-eager, interpreting instructions too permissively and taking actions beyond the user’s intent.
This is the difference between an assistant and an operator. An assistant produces a recommendation or draft for someone to review. An operator can send, publish, delete, purchase, update or change something in a live system. When the action is reversible and low risk, autonomy may be appropriate.
When an action affects customer data, production systems, budgets, contracts, campaigns, public content or financial records, the controls need to be much stronger.
Leadership takeaway:
Agent deployment should begin with authority, not capability. Define:
Which systems an agent can access
Which information it can read
Which actions it can take
Which actions require approval
What financial or operational limits apply
How activity is logged
How changes can be reversed
Who is accountable when something goes wrong
Use limited permissions, sandbox environments, backups and staged rollouts before connecting agents to important live systems.
6. AI is becoming the front door to customer service and decision routing
AI customer service is becoming easier to deploy and is being placed earlier in the customer journey.
Salesforce’s Agentforce Help Agent is becoming generally available in July. It is a prepackaged customer-service agent designed to work from an organisation’s existing knowledge and complete actions within connected systems. Salesforce is also introducing pay-per-resolution pricing, linking part of the cost to successfully resolved customer issues.
The NHS is adopting a similar principle at a much larger and more sensitive scale. A new AI triage capability within the NHS App will ask patients questions and direct them towards an appropriate service. The NHS says the tool is expected to reach more than 200,000 patients over the next 12 months and become available to all NHS App users by April 2028.
These examples demonstrate AI becoming the first point of contact rather than a tool operating behind an employee. That can reduce waiting times, make information easier to access and help organisations route demand more effectively.
However, it also gives the AI significant influence over the customer experience. A poor answer is frustrating. Incorrect routing can be more serious. An agent may fail to recognise urgency, emotion, vulnerability, accessibility requirements or a situation that falls outside its knowledge.
The quality of the experience therefore depends on far more than the model. It depends on:
The accuracy of the underlying knowledge
The quality of integrations
The actions the agent can complete
Escalation routes
Accessibility
Monitoring
Human availability
How unusual and sensitive cases are handled
Leadership takeaway:
Measure successful customer outcomes, not simply the volume of contacts diverted away from people.
Track resolution quality, repeat contact, escalation, customer satisfaction, complaints and the effect on vulnerable or high-value customers.
The objective is not to keep people away from employees. It is to resolve simple needs effectively while making it easier for customers to reach a person when judgement, reassurance or discretion is required.
7. AI legal risk starts before the prompt is written
Many organisations still associate AI legal risk mainly with inaccurate output or copyright concerns around generated content.
July’s legal developments show that the exposure is wider. A group including Hachette, Cengage, Elsevier and author Scott Turow has filed a lawsuit alleging that Google used copyrighted books to train Gemini without the necessary permission. The allegations include claims relating to books originally supplied for other services and the removal or alteration of copyright information. Google has not yet had the allegations determined against it.
Apple has separately filed a lawsuit against OpenAI and two former employees, alleging the misappropriation of confidential information connected to Apple’s hardware designs and manufacturing processes. OpenAI denies seeking or using stolen information. Together, the cases illustrate that AI-related risk can arise through:
How training or reference content was obtained
What contractual permissions were originally granted
Information employees access before changing jobs
Data passed between partners and suppliers
Material entered into AI platforms
The recruitment and offboarding of specialist employees
Ownership of the systems and processes being developed
These are allegations rather than final legal findings. But the underlying leadership considerations are already relevant.
Leadership takeaway:
AI governance should cover the full information lifecycle.
Review the origin and permitted use of data, content and intellectual property. Strengthen confidentiality, recruitment and offboarding procedures for employees working with sensitive information.
Supplier agreements should also clarify data usage, retention, ownership, liability and responsibility for third-party AI models.
The risk may exist before an employee enters a prompt or the AI produces an answer.
8. The workforce impact is becoming organisational redesign, not simple replacement
The relationship between AI and employment is becoming more visible, but it remains more complex than replacing a person with a tool.
Microsoft announced plans to cut approximately 4,800 roles while continuing to invest heavily in AI infrastructure and use AI to improve efficiency across its business. Microsoft said the reductions were not simply the direct replacement of employees by AI, although AI is changing how work is organised and where investment is directed.
Thomson Reuters is also cutting a number of engineering positions while planning to create more than 250 new roles over the next two years, particularly in senior and AI-related areas.
This points towards a reshaping of work rather than a uniform reduction in people.
Some repetitive tasks will be automated. Some roles will become more productive. Some layers of coordination may reduce. New specialist roles will appear, and judgement, relationship-building and accountability may become more valuable.
For people-led businesses, an indiscriminate headcount target can miss the larger opportunity.
AI might allow a consultancy, agency, retailer or service organisation to:
Serve more customers
Improve response times
Increase quality
Reduce administrative work
Improve utilisation
Give experienced people more time for higher-value work
Scale expertise without scaling every cost at the same rate
But those gains only appear when roles, workflows and incentives are redesigned.
Leadership takeaway:
Map the work before changing the workforce. Identify which tasks should be:
Automated
Accelerated
Improved through better information
Reviewed by a person
Kept fully human
Then consider the skills, management structure and capacity required. The question should not only be how many roles AI could remove. It should be how the organisation can create more value with the combined strengths of people and technology.
9. There will not be one global AI market
AI platforms may be global, but regulation, infrastructure and commercial access are increasingly regional.
The European Commission has issued requirements for Google to open parts of Android to competing AI assistants and share certain anonymised search data with eligible rivals.
The changes are intended to give alternative assistants access to functions available to Google’s own services. Search-data access is expected from January 2027, with broader Android interoperability requirements following in July 2027. Google has raised privacy and security concerns about the requirements.
In China, Apple’s approach is developing differently. Apple Intelligence has reportedly received approval to launch using local technology partners, including Alibaba’s Qwen and Baidu, because international models and services cannot simply be deployed in the same way as in Western markets.
This creates a more fragmented environment. A multinational business may encounter different:
Approved AI models
Data requirements
Privacy expectations
Platform integrations
Advertising environments
Search experiences
Hosting arrangements
Customer features
Disclosure and governance rules
A global organisation may therefore be unable to rely on one provider, one technical architecture or one customer experience everywhere.
Leadership takeaway:
AI strategy should be designed for portability and regional variation.
Understand where key suppliers operate, where information is processed and which regulatory or commercial dependencies could interrupt a service.
International businesses should plan AI architecture, data arrangements and customer experiences market by market, while maintaining common principles around security, quality and accountability.
10. AI’s physical infrastructure is becoming a business constraint
AI can feel like an entirely digital service, but its growth depends on physical infrastructure. Data centres require substantial electricity, water, land, equipment and connection to energy grids. Governments and local communities are becoming more concerned about who carries the cost.
New York has introduced a one-year moratorium on new data centres using 50 megawatts or more of power. Other regions, including parts of Europe, Australia and the United States, have introduced or considered restrictions because of pressure on electricity supplies, water availability, land and local communities.
Major technology companies have also signed a voluntary pledge intended to prevent the cost of AI-related grid investment from being passed to existing electricity customers.
For most marketing and business leaders, data-centre planning will not be a direct responsibility. But the consequences may still affect:
AI usage and token costs
Cloud pricing
Service availability
Regional hosting choices
Supplier concentration
Sustainability commitments
Hardware and technology costs
The pace at which AI services can scale
Businesses should also be cautious when making environmental claims about AI-driven efficiency. Removing a visible manual process does not mean the underlying technology has no physical cost.
Leadership takeaway:
AI business cases should include resilience and infrastructure exposure, not only software licences and potential salary savings.
Consider where services are hosted, how dependent the business is on individual providers and what happens if capacity, pricing or regional availability changes.
For organisations with environmental commitments, AI usage should also form part of sustainability measurement and supplier evaluation.
Final Thought
July’s developments show AI becoming more influential, more connected and more autonomous.
It is increasingly sitting between a customer and a decision, an employee and a system, a brand and its audience, or a person and the information they receive.
That creates real commercial opportunity. AI can make businesses easier to discover, improve customer service, reduce friction, increase capacity and enable people to focus on more valuable work. But an intermediary also influences the outcome.
It may decide which brands appear, which sources are trusted, which customer receives which answer and which action is taken next. The businesses that benefit most will not be those that automate everything first. They will be the organisations that are clear about:
Where AI genuinely improves the experience
Where it creates measurable commercial value
What authority it is given
How its activity is monitored
When a person must remain involved
Who is accountable for the outcome
AI is becoming more capable of acting on behalf of businesses.
Leadership now needs to become equally capable of deciding where it should.
Leader’s Checklist
A few useful questions for leadership and marketing teams this month:
Do we know how our brand appears across ChatGPT, Gemini, Google AI Mode and other AI environments?
Are we tracking both brand mentions and the sources AI systems cite?
Is our product and service information complete enough for conversational discovery?
Have we reviewed our imagery, video and structured data for visual and multimodal search?
Do we understand which AI-generated advertising and creative assets require disclosure?
Do we have permission to use every asset, face, voice and style included in AI production?
Which agents currently have access to business systems or sensitive information?
What actions can those agents take without approval?
Can actions be audited and reversed?
Are customer-service agents measured by successful outcomes or simply contact reduction?
Do escalation routes work for complex, emotional or vulnerable customers?
Have we mapped how AI will change tasks and skills before making workforce decisions?
Do contracts and employee processes protect confidential information used in AI development?
Could regional regulation or provider availability disrupt our AI plans?
Does our AI business case include usage, infrastructure, governance and environmental costs?
Useful links:
Similarweb: Did AI Take a Summer Vacation?
https://www.linkedin.com/pulse/did-ai-take-summer-vacation-similarweb-cukgf/
Semrush: Expanded 2026 AI Visibility Index, analysing 126 million AI search prompts
https://www.semrush.com/news/463141-semrush-releases-expanded-2026-ai-visibility-index-analyzing-126-million-ai-search-prompts/
Google Search Central: Introducing Search Generative AI performance reports in Search Console
https://developers.google.com/search/blog/2026/06/gen-ai-performance-reports
Google Merchant Center: Merchant Center announcements change log
https://support.google.com/merchants/announcements/6192467?hl=en
Search Engine Land: OpenAI says ChatGPT ad dismissals have dropped 50% as relevance improves
https://searchengineland.com/openai-says-chatgpt-ad-dismissals-have-dropped-50-as-relevance-improves-480991
TechCrunch: Google Images gets a Pinterest-like redesign focused on discovery
https://techcrunch.com/2026/07/14/google-images-gets-a-pinterest-like-redesign-focused-on-discovery/
Spotify: A more personal way to ask, discover and listen
https://newsroom.spotify.com/2026-07-14/talk-to-spotify-announcement-beta/
Reuters: Meta scraps AI image feature following privacy backlash
https://www.reuters.com/technology/meta-discontinues-ai-image-feature-days-after-launch-2026-07-10/
Google: Expanding AI transparency in ads
https://blog.google/products/ads-commerce/google-ads-ai-transparency-labels/
OpenAI: GPT-5.6, frontier intelligence that scales with your ambition
https://openai.com/index/gpt-5-6/
TechCrunch: OpenAI’s new flagship model deletes files on its own, users warn
https://techcrunch.com/2026/07/14/openais-new-flagship-model-deletes-files-on-its-own-people-keep-warning/
Salesforce: Salesforce announces prepackaged Agentforce Help Agent
https://www.salesforce.com/news/stories/agentforce-help-agent-announcement/
NHS England: NHS accelerates artificial intelligence rollout to cut waiting times and improve care
https://www.england.nhs.uk/2026/07/nhs-accelerates-artificial-intelligence-rollout-to-cut-waiting-times-and-improve-care-for-millions/
TechCrunch: Google faces another AI training lawsuit from major publishers
https://techcrunch.com/2026/07/14/google-faces-another-ai-training-lawsuit-from-major-publishers/
Reuters: Apple sues OpenAI and two former employees over alleged trade-secret theft
https://www.reuters.com/legal/litigation/apple-sues-openai-alleging-misappropriation-trade-secrets-court-records-show-2026-07-10/
Reuters: Microsoft joins AI-driven technology layoff wave with 4,800 job cuts
https://www.reuters.com/business/world-at-work/microsoft-joins-ai-driven-tech-layoff-wave-with-4800-job-cuts-2026-07-06/
Reuters: Thomson Reuters to cut a small number of engineering jobs
https://www.reuters.com/legal/litigation/thomson-reuters-cut-small-number-engineering-jobs-2026-07-13/
Reuters: Google required to open services to AI and search rivals under EU-mandated changes
https://www.reuters.com/world/google-required-open-up-ai-search-engine-rivals-under-eu-mandated-changes-2026-07-16/
TechCrunch: Apple Intelligence approved for launch in China with Alibaba’s Qwen AI
https://techcrunch.com/2026/07/16/apple-intelligence-approved-for-launch-in-china-with-alibabas-qwen-ai/
Reuters: Where authorities are restricting data centres amid the AI boom
https://www.reuters.com/legal/litigation/where-authorities-are-restricting-data-centres-amid-ai-boom-2026-07-14/
Reuters: New York issues moratorium on large data centres
https://www.reuters.com/sustainability/new-york-issues-moratorium-data-centers-2026-07-16/
Reuters: White House to rally utilities and data centres around an AI power-cost pledge
https://www.reuters.com/legal/litigation/white-house-rally-utilities-data-centers-over-ai-power-costs-2026-07-13/
