
AI in Insurance: what’s changing for website content
Delphine Morisset
In the insurance industry, artificial intelligence is often associated with claims management, fraud detection, and risk-based pricing. Yet AI also has a role to play in another area: managing insurers' websites and the content they publish for customers and prospects.
For insurance companies, mutual insurers, and employee benefits providers, the challenge goes beyond generating text. How can they ensure that information remains accurate across multiple websites? How can they maintain consistency in policy coverage and terms? And where should they start when introducing AI without compromising content quality?
These questions were at the heart of the "Live Business Case" panel discussion held on November 17, 2026, at Argus Factory, the French insurance industry conference hosted at Métal 57 in Boulogne-Billancourt, near Paris.
Alongside Béatrice Willems from Thélem assurances and Saber Trabelsi from Groupe AGRICA, Julian Maurel, CEO of Jahia, helped build a business case for introducing AI into an insurer's website content management processes. The exercise took place live, with audience voting at each stage, and explored the benefits, costs, and practical considerations of such a project.
Here are four key takeaways from his contribution.
Key takeaways
- AI assistants are changing how prospects discover and compare insurance providers. Brand visibility can no longer be measured solely through website visits.
- Content consistency is becoming increasingly important. Contradictory information across different pages can affect the quality of answers generated by AI assistants.
- The main investment in an AI project is not necessarily the technology itself. Documenting internal knowledge, editorial guidelines, and approval processes can require considerably more effort.
- Automating content reviews and repetitive tasks can be a better starting point than generating large volumes of content.
1. AI assistants are changing how customers discover insurance providers
An insurance company may still manage its website without using artificial intelligence. That does not mean AI is absent from its visitors' journeys.
This was one of Julian Maurel's first observations during the panel discussion.
Traditionally, marketing teams could analyze much of the customer acquisition journey through website visits, page views, and user interactions.
With assistants such as ChatGPT, Gemini, and Perplexity, part of that research process can now take place outside the insurer's website.
A prospective customer might ask an AI assistant to compare home insurance policies, explain differences in coverage, or identify products that match their needs.
The assistant may draw on several sources to generate its response. As a result, the prospect can discover different insurers, compare their offerings, and narrow down their options before visiting any of their websites.
For marketing teams, this creates a measurement challenge.
A brand may appear in an AI-generated answer without that exposure immediately translating into a measurable website visit.
Marketing teams therefore need to understand how their brand is represented in these new discovery journeys, even though conventional analytics tools still provide limited visibility into such interactions.
"We need to find a way to keep producing content and improving its quality while having access to much less data than before. That's a paradox we'll have to solve," Julian Maurel told the audience.
This development does not make corporate websites less important. On the contrary, it reinforces the need to publish reliable, accessible, and understandable information for both human visitors and the systems retrieving that information.
For a closer look at this topic, watch Jahia's webinar, "Your Next Website Visitor Is an AI", which explores how brands appear in AI-generated answers.
2. Content consistency is becoming a strategic issue for insurers
An insurance website may contain hundreds or even thousands of pages covering policies, benefits, exclusions, regulatory information, FAQs, and customer guidance.
Some organizations also manage multiple brands, dedicated portals for brokers and partners, and several language versions.
Keeping all this information consistent already represents a significant workload for content teams.
The growing use of AI assistants adds another dimension to this challenge.
Contradictory information can undermine customer trust
Consider a home insurance provider.
Its product page describes the coverage included in a particular policy. However, an FAQ published elsewhere on the website still contains an outdated version of the policy's terms.
A visitor who only reads the product page may never notice the discrepancy.
An AI assistant capable of retrieving information from several sources, however, may encounter both versions during its research.
Depending on the sources it selects and how it generates its response, it could reproduce outdated information, provide an inconsistent answer, or highlight the contradiction.
In each case, the insurer has less control over the information communicated to the prospect.
Julian Maurel describes this problem as "content dissonance."
"AI will have an overall understanding of what you're saying about a topic across multiple pages. If there are inconsistencies or contradictions, including between different language versions of your website, it will bring them to your prospects' attention."
His observation highlights the importance of managing website content as a coherent body of information rather than a collection of independent pages.
Reliable content matters more than maintaining an ever-growing library
According to Julian Maurel, improving visibility in AI assistants involves providing detailed, up-to-date, and consistent information.
"We really need to get used to continuously updating our content, even if that means reducing the amount of content we maintain," he recommends.
This does not mean systematically deleting older content or assuming that AI assistants always favor the most recently published pages.
Instead, organizations need to identify which information should be updated, expanded, consolidated, or removed when it is no longer relevant.
For insurers, this may involve changes to policy coverage, contractual terms, regulatory information, or duplicated content across several websites.
It also requires clear editorial governance: who creates information, who approves it, and how updates are reflected across different channels.
This is one of the challenges addressed by a CMS designed for insurance companies: helping teams manage content and digital experiences at scale while maintaining controlled publishing processes.
3. The real cost of an AI project: documenting what your organization already knows
When a marketing department considers introducing artificial intelligence into its editorial processes, cost quickly becomes a central question.
How much do AI models cost? What budget should be allocated to content generation? How long will it take to see a return on investment?
For Julian Maurel, these questions do not tell the whole story.
"The costs won't necessarily come from where you expect them," he warned during the panel discussion.
In his view, the cost of generating text is relatively small compared with the overall budget of a marketing team.
The real investment lies elsewhere: in the work required to communicate the organization's knowledge and rules to AI systems.
AI does not automatically understand your brand guidelines
Within an organization, many editorial rules are followed every day without ever being formally documented.
An experienced employee knows which terms to use when describing insurance coverage, which expressions to avoid, and when legal approval is required.
This knowledge often comes from experience, established working practices, and collaboration between departments.
An AI system does not automatically have access to that context.
If asked simply to write a page about home insurance, it may produce fluent text that fails to reflect the company's terminology, positioning, or regulatory constraints.
Producing usable results therefore requires clear and detailed instructions.
"You'll have to conceptualize and document everything that's implicit in your brand: your tone of voice, your brand vocabulary, how you talk about others, and the legal rules that apply when reviewing content."
A project that extends beyond the marketing department
For an insurance company, documenting these rules may involve several departments.
Marketing teams define the tone of voice, terminology, and editorial standards.
Legal and compliance teams specify the requirements governing customer communications.
Product specialists contribute their knowledge of policies, coverage, and commercial terms.
IT teams address integration, security, and technical governance.
The next step is to organize this knowledge so that it can be used within AI-assisted workflows.
This is why Julian Maurel believes that much of the cost of implementing an AI project comes from human effort and the consulting or support needed to structure that knowledge.
His analysis highlights an important principle: automation does not eliminate the need to define business rules. It makes documenting those rules essential.
4. Why automate content reviews before content creation?
Content generation is one of the most visible applications of artificial intelligence.
Yet Julian Maurel recommends that organizations consider starting elsewhere.
His advice is to automate verification processes and repetitive tasks first.
Content review as a practical first use case
Creating content involves several skills: understanding the audience, selecting an angle, developing an argument, and making complex information accessible.
Reviewing content involves a different set of activities.
These include checking whether a text follows brand guidelines, uses the appropriate terminology, and avoids potentially problematic wording.
Some of these checks can be supported by AI, provided that the underlying rules have been clearly documented.
In insurance, for example, an assistant could flag terminology that conflicts with editorial guidelines, identify wording that requires legal review, or detect inconsistencies between two descriptions of the same coverage.
Qualified employees remain responsible for final approval, particularly when content has contractual or regulatory implications.
This approach offers two potential benefits.
First, it reduces the time spent on repetitive checks.
Second, it allows employees to receive initial feedback on their work more quickly.
"Automating validation before production also helps shorten the feedback loop," Julian Maurel explains.
Authors can address certain issues before involving colleagues in reviews that genuinely require their expertise.
Automating technical CMS tasks
The same principle applies to everyday website management.
Completing metadata, checking image alternative text, identifying missing fields, reviewing certain accessibility requirements, or preparing content updates can take considerable time, especially in organizations managing large volumes of content.
"We're entering data, filling in tags, optimizing fields, improving accessibility and SEO. These are highly mechanical, tedious tasks that nobody enjoys doing," Julian Maurel observes.
AI can assist teams with these activities, provided that the results are reviewed according to the nature and risk level of each task.
This approach is also reflected in the development of agentic AI for CMS content management, where AI assistants can support content creation, review, SEO optimization, and content updates.
The objective is not simply to produce more content. It is also to spend less time on repetitive operations and more time on editorial decisions.
5. AI adoption also depends on the framework provided to employees
Introducing artificial intelligence is not solely a technology decision.
Organizations must also define which uses are permitted, who is responsible for what, and under which conditions employees can use AI tools.
During the panel discussion, Julian Maurel highlighted a gap between employees' interest in AI and the opportunities they are actually given to use it.
"Teams are very eager to work with AI, but they're not necessarily allowed to, because of concerns around sovereignty, costs, and support."
For insurance companies, these considerations are particularly important.
Organizations need to determine what information can be shared with an AI model, which solutions meet their security requirements, and how AI-assisted actions can be monitored and controlled.
They must also establish clear responsibilities.
An assistant may prepare an update, suggest a translation, or flag an inconsistency. That does not mean it should be allowed to publish information that could create contractual or regulatory obligations without appropriate approval.
According to Julian Maurel, starting with use cases that genuinely simplify employees' daily work can also encourage adoption.
A team that sees AI helping with tedious verification tasks has a concrete reason to use it.
The organization can then explore additional applications, including content generation, with established rules and processes already in place.
6. What insurers can learn from this approach
The Argus Factory panel highlighted an important distinction: introducing AI into content management is not simply a matter of adding a text-generation tool.
For insurers, the process begins with gaining better control over their existing content.
This means identifying outdated or contradictory information, documenting communication guidelines, clarifying responsibilities, and selecting tasks that can be automated without compromising accuracy.
AI can then help improve content quality and simplify maintenance while leaving decisions requiring professional judgment in the hands of employees.
This approach also addresses a broader challenge: maintaining consistent information across an insurer's digital touchpoints.
As AI assistants become part of how customers research and compare insurance products, the ability to publish reliable, structured, and consistent information becomes increasingly relevant to both marketing and IT leaders.
The first priority, therefore, may not be to generate more content with AI.
It may be to gain better control over the content that already exists.
Explore further: Your next website visitor is an AI
How do AI assistants represent your brand? Which sources do they use to generate answers? And how can you improve your content so that AI systems can better understand and reference it?
Explore Jahia's webinar on brand visibility in AI assistants and the practices that can help organizations structure their content more effectively.
Watch the webinar replay: "Your Next Website Visitor Is an AI"
Discover








