As Seen by AI: Webinar Replay and Edited Transcript

Intentful Insights Team
October 22, 2025 at 12:00 PM
18 min read

This page presents an edited transcript of the webinar As Seen by AI, led by Marina Petrova, CEO and Co-Founder of Intentful. The session recording appears on the original page. The transcript has been adjusted lightly for readability and concision. The explanations and observations in this transcript reflect Intentful's best understanding of current AI system behavior as of October 2025. While no one truly knows the exact internal mechanisms of the AI systems, the summary is based on observed behavior and hands-on testing.

Marina Petrova opened the session by framing the topic as a look at how AI systems understand websites, and more importantly, how those systems interpret brands.

The webinar focused on three central areas. First, it covered what matters now, in October 2025. Second, Marina walked through examples of how AI reads selected websites submitted by volunteers for review. Third, she demonstrated agentic search and explained how companies and destinations can run a similar exercise with their own teams to understand how AI interprets their web presence. Based on those observations, she then shared recommendations that Intentful considers important for discoverability.

The larger shift is not only about search. AI is altering the way customers, visitors, and users interact with information and content.

For attendees less familiar with Intentful, Marina explained that Intentful started in 2021 and has been building AI that understands brands, destinations, and companies since its founding. The company works with organizations around the world across tourism and travel, performing arts, CPG/FMCG, telecommunications, agencies, and other sectors as its global work continues to grow.

The guidance in the webinar comes from Intentful's daily experience building with AI. Intentful's products are powered by AI search, so the recommendations are grounded in practical work rather than ideas gathered from online articles or thought leadership commentary. Because AI behavior keeps evolving, the knowledge also continues to change.

Grounded in Daily AI Work, Not Abstract Theory

Although search was discussed throughout the session, Marina emphasized that the change underway is much broader than search alone. It is a change in how information is accessed, processed, and used. The familiar points about search disruption and declining traffic are already widely understood; the webinar instead concentrated on what brands can do now.

Each AI Response Draws on Hundreds of Sources

In the previous model of optimization, most work centered on a company or brand website. Today, when an AI system answers a single query, it may evaluate hundreds of information sources, and sometimes even more. That means optimizing a web presence is no longer only about improving the main website. It is about the many places where a brand may appear or be referenced.

Marina noted that one of Google founders has suggested the number may be thousands {in a single query analysis}. Whether that was meant literally or as an illustration of scale, Intentful would not be surprised if AI systems already evaluate thousands of sources before returning an answer.

That creates a deeper challenge than website optimization alone. If a brand does not provide clear input, AI will form its understanding from what it can find rather than from what the brand intends to communicate. This matters across content programs, PR campaigns, partner listings, local profiles, and any other source where the brand is mentioned.

At a simplified level, brands now have two major goals. The first is to be discoverable to AI systems. The second is to influence and own the story AI tells about the brand.

Goal One: Being Discoverable

While preparing for the webinar, Intentful corresponded with companies and destinations that volunteered their websites for analysis. One destination and its agency later asked not to be used as an example because AI's description of the destination did not align with how they wanted to be represented. That illustrates the amount of work required to influence the story AI tells.

Marina emphasized that content remains extremely important, but it is not the first step. Content becomes the second goal. The first goal is discoverability.

Most websites are not yet optimized for AI systems because they were built for a different time and a different playbook. Their structure often makes important content invisible to AI. In the demo, this became clear: pages that look complete to humans can appear incomplete or empty to AI systems.

There are three immediate discoverability considerations.

First, a website must be open to AI bots and AI crawlers. Some organizations hesitate to allow AI bots because of concerns about hosting costs and because AI crawlers behave differently from Google. Those concerns are real, but they can be managed. Brands can set limits around how often bots visit, but the first requirement is allowing them access. Without access, those systems cannot include the site in their knowledge bases.

Among the websites submitted by volunteers, some blocked roughly half of the AI bots. For companies based in the EU, Marina noted that some websites were entirely blocked from AI systems, meaning the systems reached the website but were not allowed inside. In that case, even strong, polished content cannot be seen by AI.

When Marina referred to AI, she meant bots from OpenAI, Perplexity, Anthropic, and Google, among others. Traditional SEO remains important for several reasons, but discoverability must now include AI systems as well as traditional search.

Every website now serves two audiences: the human audience and the AI system audience.

How AI Reads a Website Compared with How People See It

For the examples, Intentful used volunteers from Destination Marketing Organizations, or DMOs, which promote destinations to visitors and support local communities. Marina described DMO websites as visually rich, thoughtful, and carefully built, often with strong imagery, color, style, and a clear sense of place.

However, those sites now need to work for two audiences. The webinar included screenshots from volunteer websites that looked colorful, polished, and well organized to human visitors. They contained natural navigation, visual structure, and clear information for people.

One example focused on Oconomowoc in Wisconsin. Marina noted concern about pronouncing Oconomowoc correctly and described the website as typical of many destination websites. It includes information about things to do, places to stay, restaurants, and related visitor content. For people, it appears inviting, with an embedded live video and an inspirational design.

That is how a person sees the homepage. The AI system, however, sees the same page very differently. What appears complete to humans is not necessarily what an AI system can access.

Another page, Things to Do, contained valuable destination information: beaches, fishing, water activities, boat rentals, theater, concerts, and sports. A significant amount of work had clearly gone into collecting, organizing, and presenting that content in a usable way. But when the AI system read the same page, it surfaced only a portion of what was actually there.

Compared with many other DMO websites, that result was still relatively good. Some sites show AI only one paragraph or a few lines, while the rest of the human-visible information is missing. If an AI bot reaches a page and cannot find useful information in seconds or milliseconds, it moves to another source. If it cannot see the content it needs, the brand loses that opportunity to connect with a potential visitor.

In another randomly selected section named Water, the AI system could see only the page title. The rest appeared as a blank page.

This was not unique to one destination. Across multiple submitted websites and destinations, the pattern repeated: beyond the header, AI often found little or no substantive content.

Discoverability matters because it ensures the work, budget, and creative effort invested in content can actually be used by AI systems. Without it, content remains available only to human visitors. Advertising can still drive traffic, but for organic visibility, pages cannot appear empty to AI.

Agentic Search Demo in ChatGPT

Marina then demonstrated Agentic Search in ChatGPT, using a pre-recorded example created the day before the webinar so the information would be current.

The query was fictional. Marina was not actually speaking at the conference referenced in the prompt, but because she used her ChatGPT Business account, the system already had context about her. The prompt asked for recommendations for where to stay in a destination she had never visited, along with suggestions for the best places to eat and other visit-related guidance.

A key instruction was to activate agentic mode rather than relying on a standard ChatGPT search. Agentic mode makes it possible to observe the system's reasoning and decision process, which offers insight into how it chooses sources, follows links, abandons pages, and composes an answer.

The destination used in the demo was Oconomowoc in Wisconsin. Marina encouraged attendees to try the process themselves, ideally recording the session so they could pause and review the reasoning step by step.

What the Agentic Demo Revealed

The agentic search showed how the model moved through sources, evaluated access, and synthesized information:

• It first chose to open the official DMO website, treating it as a credible local source for lodging information beyond TripAdvisor. This was notable because the model distinguished between official and unofficial sources, which earlier versions did not do as clearly.

• It reviewed the site's dining section, found that the main page showed restaurant names without much additional detail, and then followed restaurant links to seek fuller listings.

• When the website did not provide enough information, the model returned to search results and opened individual restaurant websites.

• On restaurant sites, it assessed available information and recognized that even simple pages could contain useful context in sections such as Our Story and Local Partners.

• The search process moved much faster than older versions, taking about 5 minutes rather than 10–15, while still collecting and combining details from several websites.

• The model then visited Travel Wisconsin, the state tourism website, but identified that automated access was blocked. It quickly stopped trying that source and moved to accessible sites including Twisted Fire.

• On the Twisted Fire site, it observed that heavy image use reduced readable text. That mattered because content is essential, but crawlability and readable structure come first.

• It attempted to use MapQuest and other sources, while occasionally returning to Travel Wisconsin despite the access block, showing persistence in its reasoning process.

• When it visited Yelp, dynamic loading prevented content extraction. The model recognized the issue, left Yelp, and continued to Badger Burger's own website.

• On the Badger Burger site, it reached a long page, navigated to code line 888, and processed the content efficiently. This demonstrated how quickly AI can scan and interpret extensive page content.

• It evaluated text such as flavorful burger ingredients, judged whether the information would help travelers, and returned to the DMO site for additional context.

• The overall process resembled human research behavior: exploring, checking sources against one another, and combining findings. The difference is that AI can do this at a scale and speed humans cannot match.

• In the end, the model generated a coherent summary using material from the DMO, restaurant, and travel sources.

Running a recorded agentic search for a destination or business allows teams to see how AI understands their online presence, including what it can read, what it ignores, and which sources it tends to prioritize.

Marina did not ask follow-up questions in the demo because the purpose was to show how the model searches. In a real user journey, a person would probably continue with more specific questions about their trip.

She strongly encouraged attendees to watch how AI evaluates their company websites and to note which external sources it uses.

Three Main Steps for Getting Your Website Into the AI Source Set

Marina then outlined the steps brands can take to improve the likelihood that their website appears among the sources AI uses.

The framework has three core steps. Some attendees had heard Marina discuss them before, and she emphasized that they remain relevant.

Step 1: Discoverability

The first step should become a checklist for discussion with the web team or developers. Teams should review each point and ask for specific confirmation about whether the site is open to Perplexity, ChatGPT, Google, Apple, AI, and others. If access is restricted, the team should understand how often bots are allowed in, such as once a day or three times a day, so they can collect information.

For sites with significant traffic and concerns about hosting costs, the team should review current robot.txt rules and what those rules allow. A development team can show this clearly, and a non-technical stakeholder can still understand whether the setup appears reasonable.

Sitemap quality deserves particular attention. Marina said that, in 95% of cases, sitemaps are a mess.

The first requirement is having a sitemap at all. In the websites Intentful has reviewed, probably 35% do not even have a sitemap.

Once the sitemap is accessible online, teams should spend about an hour reviewing which pages are included. The sitemap is not merely a technical file; it is a map that helps AI understand what content exists on the website. Often it contains outdated material, irrelevant pages, and content that no longer reflects the site. In the older search environment, keeping old pages sometimes made sense because a visitor might land on one and continue deeper into the site. That logic is weaker now, so sitemap content should be current and relevant.

Teams also need to confirm that all desired pages are actually present in the sitemap. Intentful often sees published pages that are not included. Unless traffic is driven to them through ads, those pages may never be discovered. Paid traffic is one option, but organic discovery should not be overlooked.

After reviewing a first sitemap, the process becomes easier to understand. Teams should clean it up, keep it current, and clarify whether sitemap updates are manual or automated.

Accessibility also matters. In this context, Marina referred to accessibility from the accessibility standpoint: the website should be optimized according to applicable standards, or as close to that as the team can manage. This also affects what AI can see.

The ChatGPT agentic reasoning demo showed an example involving dynamic rendering. Marina did not suggest removing dynamic loading entirely, since that may not be practical. Instead, she recommended discussing with developers whether dynamic elements can also be exposed in a static form. AI needs to understand that the content exists. If an agent is looking for tickets, hotels, or dates, it may click into dynamic elements, but the page must first make clear that meaningful content is present rather than appearing blank.

Step 2: Structure and Signals

The second step concerns structure and signals. Much of this comes from traditional search and basic SEO hygiene.

If a page takes too long to load, AI may skip it and choose another source. PageSpeed Insights is one tool non-technical teams can use to check whether pages meet speed and timing guidelines. If performance is not within guidelines, the next step is to ask developers what can be improved. Marina connected this to the ChatGPT demo, where the model effectively decided not to load content because images were too heavy.

Basic SEO tags still matter. Open graph tags still matter. Structured data is especially important because it helps AI understand and read the page. This applies not only to ChatGPT or Perplexity but also to Google, where structured data has long been important and will remain so.

Structured data is as important as the sitemap. Based on the companies registered for the webinar, Marina noted that many had events or other content types that require structured data. For anyone unfamiliar with structured data, Google provides documentation on available types, and web teams can usually implement it without difficulty.

Semantic markup is also important. Teams should also avoid unnecessary pagination that hides content behind extra clicks.

Step 3: Content Interpretation

The third step is content.

Content remains extremely important. Keywords are no longer treated the way they were in traditional search, but they still matter as signals and should not be ignored completely.

Text content is necessary. AI can read images, view images, and watch videos, but text is still the first point of entry. Pages should include descriptive content about the topic they are meant to cover. That content should be clear and informational, not only marketing language.

Freshness is also important. AI systems can check recency, and they often try to ensure the material they use is current rather than relying on references from 2016, a pattern Intentful still sees from time to time.

The context window will keep expanding. Previously, AI considered a shorter amount of text when interpreting a topic. Now it can include a much larger volume, approaching book-scale amounts of text in simplified terms.

This creates two content streams. Brands should absolutely continue producing inspirational content for humans. People still want imagery, video, color, atmosphere, and emotional connection.

At the same time, both humans and machines need informational content.

That informational content must be specific. If the brand knows something but has not placed it on the website, AI will not know it. AI also makes it possible to move beyond assumptions about what users might want and instead examine actual intent.

Marina said she believes this is an extraordinary time because brands now have the opportunity to connect with every customer, user, and visitor in ways that were not possible before.

A few months before this webinar, Intentful hosted another session that examined a sample of 15,000 questions from destination websites. Those were questions people asked through the Intentful AI Assistant installed on destination websites. Marina did not repeat that full webinar, but she pointed to it as a strong example of how teams can learn what people are genuinely interested in and what they ask about.

There is no PII attached, and Intentful does not collect personal information in that context. The data is anonymized, so Intentful does not know who the people are. Even so, it provides deep insight into what visitors seek when they arrive on a website.

For DMOs especially, Marina encouraged teams to ask whether their websites contain enough information for AI to answer visitor questions. That applies whether the AI assistant is installed on the website, a ChatGPT AI bot visits the site, or Google crawls the content.

In the DMO context, people interact with AI as though they are asking a trusted local rather than an AI system. They ask about hours, parking, average costs, restaurants, and other practical details. They want information, not only inspiration. As teams plan content strategy, they should be as detailed as possible. Marina acknowledged that this requires significant work, but if the goal is helping visitors engage and find answers, the content needs to be updated.

The change is also broader than search. Companies are moving from speaking at customers through websites, ads, and marketing broadcasts to speaking with customers. This is the beginning of two-way communication.

Bringing Knowledge and Content Together

Once discoverability is addressed and AI no longer sees a website as a blank page, AI works with the knowledge the brand publishes. It breaks information down into chunks today, although this will change as context windows grow. At present, it separates content into smaller manageable pieces, understands context, and reassembles relevant pieces when answering a question.

To appear in those answers, a brand needs to be discoverable and needs to make the relevant information available on its website. Professional knowledge should guide content decisions, but customer questions should also be used as input.

A content strategy for humans and machines should balance inspiration with information. People ask in-the-moment questions. Intentful sees this repeatedly: people want to know what is happening right now. Someone may be stuck in traffic and ask AI a question as if AI already knows the situation.

For DMOs, Marina particularly emphasized updating member pages. It is not enough to list only a restaurant name or an event title. Pages should include as much useful information as possible, beyond marketing language. Teams should act like trusted guides or front desk guides, not only marketers, and keep details current. Information from several years ago is increasingly unhelpful.

The Four Buckets Checklist

Marina summarized the recommendations into four buckets:

1. Discoverability comes first. Teams can see this by using the examples from the webinar or by opening ChatGPT, activating agentic mode, and observing what it knows about the brand.

2. Structure and signals need to be in place. That includes structured data, a sitemap, and related technical foundations. Teams should sit down with their web teams, who will understand many of the necessary actions. For many developers, AI search is still relatively new because they do not work with it every day. The point is not that they did something wrong; it is that they also need to engage with this new requirement.

3. Content should combine inspiration with practical information. Teams should look at real user information, make sure pages are informative and substantial in length, and keep content fresh and accurate. They should also remember the shift from speaking at customers to speaking with customers.

4. Brands should continue monitoring how AI describes them across multiple sources, not only on their own websites.

Intentful Products for Discoverability and Engagement

Intentful offers two categories of products for customers.

The first category focuses on discoverability and how AI interprets a brand or destination. The program is called As Seen by AI. Intentful recently launched it as a 12-month program and is already working with several DMOs. Marina emphasized that this is not a one-time effort because AI keeps changing and because meaningful discovery and website changes are required for AI to see the brand accurately.

The second category focuses on applying AI to user engagement. Intentful's AI Suite includes an AI assistant, generative response ads that allow people to converse with an ad and receive real-time responses, and an on-brand content tool powered by the same AI that knows the brand.

Marina closed by encouraging attendees to reach out through LinkedIn, the Intentful website, or email. She wished them luck optimizing their websites and reminded them that the environment will keep changing, while the foundations remain consistent. If a brand has already optimized for SEO, that provides a useful starting point for AI search optimization. There is more work to do now, but the starting point is discoverability: make sure AI can see the website, and take back control.

About This Article

This article is an edited transcript from the webinar As Seen by AI, covering how AI systems read and represent brands online, the gap between what humans see and what AI sees on websites, an agentic search demo in ChatGPT, and actionable recommendations for improving AI discoverability, structure, and content. Intentful's As Seen by AI© program and AI Suite products are referenced.

Intentful is commercially deployed since 2021, working with organizations in travel and tourism, performing arts, CPG, telecommunications, and agencies globally. Contact: [email protected]