Many marketing teams have spent 2026 explaining the same chart to their executives. Organic traffic is down, inbound leads have followed, and the search rankings that used to predict both have barely moved.
The cause is well documented. When Ahrefs first measured the effect in April 2025, a Google AI Overview was associated with a 34.5 per cent lower click-through rate for the top-ranked page. When it repeated the study on December 2025 data, the figure was 58 per cent. Buyers are getting their answers from ChatGPT, Gemini, Perplexity and Google’s own summaries, and many never reach a brand’s website.
Contentful calls the shift the Great Content Collapse, and while most attention so far has gone to the lost traffic, the greater exposure sits inside the answer itself. An AI assistant asked to recommend a vendor will describe each company, compare it with competitors and give the buyer a reason to choose one. It builds that description from whatever it can find, whether or not anyone at the company has checked the result.
Content now has two audiences
Traditionally, content was written for people, to inform a human reader and move them towards a decision. That has changed now, and content now needs to be written with the understanding that there will be a second reader.
AI agents sit between brands and buyers, and they reward structure, completeness, authority and context. Meanwhile, people still expect authenticity, relevance and credibility, and the two sets of expectations can, with the wrong content strategy, pull against each other.
As anyone in marketing would be happy to relate to you, copy that persuades a person can be hard for a machine to parse. What that means in the era of LLMs is that content that is engineered purely for machines might do a good job of catching their attention, but if it reads as robotic, readers will punish it. In a Raptive study of 3,000 US adults, people who suspected content was AI-generated rated it 48 per cent less trustworthy, and purchase consideration for brands advertised beside it fell 14 per cent.
Martech leaders find themselves in a position where they are asked to serve both audiences at once, and few have been handed extra resources to do it.
Who owns it
In fact, while most organisations have yet to answer this, and the lag is understandable, because the symptoms surface in different parts of the business.
SEO teams tend to notice first, since they watch the prompts and rankings. Product marketing hears a positioning gap, brand sees a reputational risk, PR sees thin third-party coverage, and the web and content teams inherit the fixes.
All of them have a point, because answer engines read far more than the marketing site. They draw on media coverage, review sites, analyst notes, partner pages and help-centre articles. An old support note about a long-resolved limitation can end up shaping how an assistant describes the product today.
Contentful’s position is that AI reputation is too consequential for any single team to own, and that those teams need a shared body of evidence to work from. Someone still has to bring them together. There is a strong case for that person being the martech lead, who already runs the content platforms, holds the analytics and sits between marketing, product and IT.
The tools and processes
Among the teams that have made a start, the operating model has three parts.
- Structure the content once. Content held as modular, structured entries in a headless CMS can be delivered to websites, apps, marketplaces and AI agents from a single source. The structure makes specifications, pricing and proof points legible to a machine, and keeps the facts consistent wherever an answer engine encounters them.
- Use AI for speed, with governance attached. AI can carry the research, localisation and reuse work so that teams spend their own time on relevance. Volume for its own sake feeds the trust problem, and approval controls matter more as content is assembled dynamically across channels.
- Benchmark the answers, then fix one gap at a time. This is the newest piece. Palmata uses a research agent called Sounder to benchmark how answer engines represent a company across the topics, audiences and competitors a team nominates. It traces the sources and content gaps behind those answers, recommends specific changes and models the likely effect of each before any budget is committed.
The process Contentful recommends is modest. Choose one commercially important question, such as a product launch or a priority segment, and run a benchmark. Bring SEO, product marketing, content, web, brand, PR and support into the same readout, agree on the gap that matters most, and give one or two actions to named owners. Repeat the benchmark once the work ships.
Making the remaining visits count
A more accurate AI reputation is one half of the recovery. The other half is converting the visitors who do arrive, and the same structured content does that work too.
One example is Pets Deli, a European direct-to-consumer pet food retailer that used Contentful Personalisation to tailor pricing and promotions by customer segment during a Black Friday campaign. Conversion rates rose 51 per cent and bounce rates fell 10 per cent.
Search traffic is unlikely to return to its old shape. What marketing teams can control is the accuracy of the story answer engines tell about them, and the experience waiting for the buyers who click through. Both begin with knowing what the machines are saying today.
Find out how AI answer engines represent your company. Sign up to try Palmata for free.
