BrightonSEO talk roundup October 2026

This October, a few of the Varn team headed back to the world’s biggest search marketing conference, brightonSEO, to hear from a variety of industry peers on their specialist topics.

This autumn’s agenda was packed with sessions on AI search, measurement, ecommerce and the skills teams need to deliver results. We’ve rounded up a few of our favourite sessions below.

Jean Dejsuvan explained why LLMs don’t behave like traditional search engines and why “AI ranking” is the wrong mental model, covering what makes a brand visible and citable to AI agents and the common mistakes draining time and budget.

Key takeaways:

The talk gave three key points for brands to focus on:

• Build a brand worth recommending
• Make your brand understandable
• Give AI evidence everywhere

Marco Bonomo demonstrated how to build Python and AI-powered Streamlit apps that automate manual work, with examples including an alt text generator and a redirect mapping tool for migrations and 404s.

Key takeaways:

• Marco was keen to state that “It’s not too late to get started with vibe coding”
• Python can help AI by giving it the ‘rules’ to work with, meaning you get the same answer every time (avoiding hallucinations)
• AI can help you make the tools you’ve always wanted.
• Marco demonstrated a few tools he’s built using Python and AI, including an automatic redirect mapping tool, and a technical SEO health check dashboard for multiple sites.

Josh Blyskal broke down how ChatGPT, Google AI Mode and Claude each choose which sources to cite, using a dataset of real AI search results built at Profound. He covered why Claude is the most predictable engine to win, what changed when ChatGPT opened its ads to self-serve, and how Google AI Mode reads a page differently depending on whether the query is about a place or a product.

Key Takeaways:  

• Josh provided plenty of interesting data points from Profound’s data set – particularly useful because Profound typically has access to a lot of AI search data, so can be ahead of data trends.
• Josh showed how Query Fan Outs nearly doubled on August 8th on ChatGPT – this means ChatGPT is ‘searching’ for more information to answer queries, and shows an evolution of the model even further.

Nikki Goodman explored why good SEO recommendations often fail to reach the roadmap, and shared a practical framework for evaluating impact, effort and confidence so teams can prioritise work that aligns with business objectives.

Key takeaways:

• SEO recommendations have to compete with other priorities, such as: compliance and leadership queries, be filtered through constraints and bias, and be objectively evaluated before it goes into the roadmap.
• Showing “SEO impact” does not always equal implementation priority. SEO managers need to show SEO value, business value and delivery reality.

Neil Barnes explained why directional data beats measurement perfection, showing how bringing disparate, “messy” data into a single view supports the pattern recognition and storytelling needed to drive business change.

Key takeaways:

• Prioritise direction over perfection.
• Simplify reporting setups by consolidating core KPIs into a single brand health scorecard or dashboard.
• Ensure time for analysis and storytelling is protected, focusing on reviewing data to generate insights and recommendations that link back to business KPIs.

Ollie Sloan showed how to structure board reporting through a commercial lens, including demand trends, efficiency and marginal returns, and how to report uncertainty honestly without cluttering reports with metrics that add volume but not clarity.

Key takeaways:

• Shift reporting from describing “What happened?” to recommending “What do we do now?”
• Use detailed dashboards internally, but present leadership with short, concise reports that only show the data needed to prompt specific decisions.

Malte Landwehr shared learnings from tracking more than one million products across more than 250 million ChatGPT conversations, covering the levers that influence which products ChatGPT recommends, including product attributes, fan-out terms, landing pages and the product feed.

Key takeaways:

• LLMs in the US now use data from their own database (Labrador) rather than Google, and pull information from product feeds
• When optimising an ecommerce website for AI shopping channels, you need to ensure all products have feed data, not just a handful
• There is a cap on product feed descriptions, so ensure the most important information is included in the first half of the description

Stijn Bergmans shared seven tactics from building an in-house AI stack across 200+ client cases, including how to measure Answer Share, why tracking streaks beats one-off visibility scores and how to find the influential citations that shape an answer.

Key takeaways:

• It is important to understand that people often use multiple channels before making a purchase. For example, they may discover products and understand which product they may need using an LLM, and then continue a more specific search in Google
• Therefore, an SEO & GEO strategy needs to focus on AI visibility as well as search engines and social media channels

Greta Koivikko shared methods for turning product detail pages into mini storefronts that get discovered across search engines and AI platforms, covering product attributes, situational relevance and agentic commerce readiness.

Key takeaways:

• Product description pages (PDPs) act as ‘mini storefronts’ for your website, so when a user lands on one, they may use the information available to decide whether to continue shopping on your website.
• Due to this, you should focus on PDPs over Product Listing Pages (PLPs), as you can present more information on PDPs.
• When optimising PDPs, follow the SEED framework: Situational relevance (target the right people), Example-led content (use scenarios and situations for products), Enrich product data (use structured data and feeds), and use a Distinct brand voice

Joe Worgan showed how to run custom extraction crawls at scale to uncover what makes a page perform in traditional search and LLMs, connecting the data to real performance results.

Key takeaways:

• Start using custom extractions to analyse the performance of specific page elements such as FAQs, content freshness etc.
• Use extracted insights to guide design and content changes.
• Extract multiple elements at once to provide bigger picture analysis.