6 November 2026 · 11 min read
AI already speaks for the company. Who decides what it says?

Key Takeaways
- In February 2024, the Civil Resolution Tribunal of British Columbia held Air Canada liable for a chatbot answer that contradicted the bereavement policy published on Air Canada's own website.
- The Future of Websites, a 2026 report by Scott Brinker (chiefmartec) with Knotch, proposes a dual-surface website: one surface for people, one for AI agents, both fed by a single brand context model.
- A brand context model is a single, current representation of a company's knowledge, offers, policies and capabilities, from which the website, the chatbot and every AI agent draw.
- Automated traffic reached 51% of all web traffic in 2024 (Imperva), and only 1% of visits to Google searches with an AI summary included a click on a cited link (Pew Research Center, 2025).
- The decisive step belongs to the board: appointing one owner for the company's official version before buying AEO services, chatbots or brand context platforms such as Sitecore's Brand Context.
In November 2022, Jake Moffatt bought a full-fare Air Canada ticket from Vancouver to Toronto to attend his grandmother's funeral. Before buying, he asked the chatbot on Air Canada's website how bereavement fares worked. The answer was clear: he could request the reduced fare afterwards, within 90 days of the ticket being issued.
When Moffatt submitted the request, Air Canada refused. The bereavement policy published on Air Canada's website said something else: bereavement fares do not allow refunds for travel that has already happened.
The case went to the Civil Resolution Tribunal of British Columbia. Air Canada argued that the correct version of the policy had always been on its website, and went as far as describing the chatbot as "a separate legal entity that is responsible for its own actions." In February 2024, tribunal member Christopher Rivers rejected the argument:
"While a chatbot has an interactive component, it is still just a part of Air Canada's website. It should be obvious to Air Canada that it is responsible for all the information on its website. It makes no difference whether the information comes from a static page or a chatbot."
The airline was ordered to pay 812 Canadian dollars in damages and fees.
The amount is irrelevant. The Civil Resolution Tribunal found one company, on one website, with two versions of the same policy, and ruled that Air Canada answers for both. The same situation exists today in most companies, and generative AI is multiplying the places where it can surface: website chatbots, AI assistants such as ChatGPT, and AI agents acting for customers.
What is the dual-surface website Scott Brinker proposes?
The dual-surface website is Scott Brinker's model for the AI-era website: one surface for people, one surface for AI agents, and a brand context model underneath that feeds both. Brinker, author of chiefmartec and one of the reference voices in martech, sets it out in The Future of Websites, a 2026 report published with Knotch.
Page debt and brand latency. Brinker starts from a diagnosis any CMO will recognise. Page debt is marketing's flavour of tech debt: every campaign, segment and journey adds pages and variations. When the business changes, the change takes time to reach all of them. Brand latency is the result: the gap between what the brand intends customers to see and what customers actually see.
Two surfaces. The Future of Websites proposes a website with:
- A surface for people, which adapts to what each visitor says they want, instead of forcing them through a page architecture designed by the company.
- A surface for AI agents, with machine-readable content, structured data and services an agent can call: check availability, calculate a quote, book a meeting.
The brand context model. A brand context model is a single, current representation of the company's knowledge, offers, policies and capabilities, from which every surface draws. Brinker sums up the architecture in one line: "The page is a rendering. The website is a lens. The brand context model is the core."
The dual-surface website, with the brand context model underneath. Original graphic by Scott Brinker (chiefmartec), figure 2 of the report "The Future of Websites", published in partnership with Knotch (2026). Reproduced with credit to the author.
51%
Of web traffic in 2024 was automated, surpassing human activity for the first time in a decade (Imperva)
1%
Of visits to Google searches with an AI summary included a click on a link cited in the summary (Pew Research Center)
The numbers explain the urgency. Automated traffic already accounts for 51% of web traffic, according to Imperva's 2025 Bad Bot Report. When Google shows an AI summary, users click a traditional search result in 8% of visits, against 15% when no summary appears, according to a July 2025 Pew Research Center analysis of 68,879 Google searches. A growing share of how a company is represented in Google, ChatGPT and other AI assistants happens without anyone opening its website. Brinker describes the effect this way: AI assistants "deliver the experience of the AI, not the experience of the brand."
What question should boards ask first?
Boards should first ask who owns the official version of the company, and how long a change in the business takes to reach every touchpoint. Most discussions start at the surface instead. How do we show up in ChatGPT's answers? Which agency should we hire for AEO? Which chatbot should go on the website? What structured data do AI agents need?
These four questions are fair, and they all assume something most companies lack: an official version of the company, current and owned by someone.
The Future of Websites describes where that version lives today. Content sits in the CMS, images in the DAM, product data in the PIM, brand rules in a wiki, campaign briefs in a shared folder, legal policies on an isolated web page. Each of these sources belongs to a different department, from marketing to product to legal, and updates at a different pace.
Architecture of a dual-surface website: company sources below, the brand context model in the middle. Original graphic by Scott Brinker (chiefmartec), figure 6 of the report "The Future of Websites", published in partnership with Knotch (2026). Reproduced with credit to the author.
Every new surface the company adds, whether a chatbot, an AI assistant or an MCP service for agents, draws on those same sources. When the CMS and the PIM contradict each other, the new surface repeats the contradiction in direct conversation with the customer, with the same legal responsibility the Civil Resolution Tribunal assigned to Air Canada.
Can a brand context platform solve the problem?
A brand context platform organises the company's truth, and it still needs someone in the company who owns that truth. Vendors saw the issue before most boards did. In August 2026, Sitecore launched Brand Context in SitecoreAI: Brand Context analyses the brand's website and organises the information into structured files that AI agents use to produce on-brand content. Other content and AI-visibility platforms are launching features with the same goal.
Sitecore's documentation includes a telling detail: only organisation admins and owners can edit Brand Context files. SitecoreAI assumes someone in the company is accountable for the truth those files hold.
CRM taught the same lesson. A CRM project fails when nobody owns the process and the board approves the budget without approving the change. A single source of truth bought as a product, with no named owner, becomes one more silo to keep in sync with the CMS, the DAM and the PIM.
Brinker has a line that captures the commitment: "It's the composition that is generated, not the ground truth." Generative AI can compose the experience around each visitor. The facts it composes from must come from a source someone governs.
How does a brand context model connect to hyperpersonalisation?
A brand context model is, in practice, the half of hyperpersonalisation that sits on the company's side. The Future of Websites names the point where everything meets golden context: the intersection of customer context (what the customer wants), company context (what the business can deliver) and systems context (what the infrastructure can execute in the moment).
Golden context: where what the customer wants, what the business can deliver and what systems can execute meet. Original graphic by Scott Brinker (chiefmartec), figure 8 of the report "The Future of Websites", published in partnership with Knotch (2026). Reproduced with credit to the author.
Most personalisation projects invested in the first circle: customer data, segments, CDPs. The second circle, company context, was rarely built. That gap is why so many personalised experiences get the customer right and the offer wrong, showing the right person the wrong price or an outdated condition. Without a brand context model, personalisation at scale multiplies errors as efficiently as it multiplies hits.
Objections worth answering
"AEO delivers results now. Organising the company's truth is a multi-year project." Optimising a website full of contradictory facts for citation in ChatGPT or Google's AI Overviews only widens the reach of the contradictions. The first step fits in a quarter: identify the 20 to 30 facts customers and agents ask about most (prices, terms, lead times, availability, return policies) and give each an owner and an update deadline. The Future of Websites draws exactly this line between what can be generated and what "must remain exact."
"This is an IT or martech matter, not a board matter." The sources behind the company's official version belong to marketing, product, legal, sales and operations. None of those five departments has authority over the others. Only the board can appoint a cross-functional owner and give them a mandate to settle contradictions.
"AI assistants form opinions from what the market says about the brand, not only from the website." This is the strongest objection, and it widens the problem rather than dismissing it. The company's official version has to be consistent on the website, in marketplaces, in directories, on partners' product listings and in support documents. The more external sources ChatGPT and other AI assistants read, the more it matters that one internal origin has an owner.
What to take to the next board meeting
- Measure brand latency. Pick a recent change (a price, a commercial term, a policy) and measure how long it took to reach the website, the chatbot, partners and ChatGPT's answers. The gap between the first and the last touchpoint is the company's brand latency, expressed in days.
- Appoint an owner for the company's official version. One person with a cross-functional mandate, accountable for the facts customers and AI agents consult, with authority to resolve contradictions between departments.
- Approve tools after the owner. Brand context platforms, chatbots and AEO projects work better when they arrive at a company that already knows who answers for the truth they will repeat.
The question that remains
AI already speaks for the company, in Google's search results, in the assistants customers use and, increasingly, on the company's own website. In Moffatt v. Air Canada, the Civil Resolution Tribunal of British Columbia made it clear that the company answers for everything said in its name, wherever it comes from.
So the most useful question to take to the board is a simple one: if a price changes tomorrow, how long until the website, the chatbot and ChatGPT all say the same thing? The answer shows who actually owns the company's official version, or that nobody does.
Sources
- Brinker, S. (2026). The Future of Websites: A Brand Context Model to Power Human & Agent Web Experiences. chiefmartec, in partnership with Knotch. https://content.martechday.com/the-future-of-websites-brand-context-model.pdf (figures 2, 6 and 8 reproduced in this article with credit to the author).
- Cerullo, M. (2024, February 19). Article on the Moffatt v. Air Canada decision. CBS News.
- Brody, J. M. (2024, February 28). AI Gone Wild: Airline Has to Honor a Refund Policy Its Chatbot Fabricated. Manatt, Phelps & Phillips.
- Moffatt v. Air Canada, 2024 BCCRT 149 (Civil Resolution Tribunal of British Columbia, February 2024).
- McKeever, G. (2025, April 15). 2025 Imperva Bad Bot Report: How AI is Supercharging the Bot Threat. Imperva.
- Chapekis, A., & Lieb, A. (2025, July 22). Google users are less likely to click on links when an AI summary appears in the results. Pew Research Center.
- Sitecore. (2026). Brand context [SitecoreAI documentation and changelog of August 31, 2026].