The AI search playbook: lessons from BrightonSEO San Diego
The AI search playbook: lessons from BrightonSEO San Diego
I've just returned from BrightonSEO San Diego, and I'm still processing everything I learned. The flight back from sunny California to cold Estonia gave me plenty of time to reflect on what was easily one of the most pivotal conferences I've attended in years.
If you've been following the digital marketing space, you'll know that AI search has been the topic on everyone's lips. But San Diego wasn't just another echo chamber of "AI is changing everything" platitudes. What I witnessed was a fundamental shift in how we need to think about search, content, and brand visibility in 2025 and beyond.
Here's what stood out, what matters, and what we're implementing at Blu Mint as a result.
The death of traditional SEO has been greatly exaggerated
Let's start with some perspective. Ahmed's opening session reminded us that people have been predicting the death of SEO since 1997. We saw it again in 2015. And here we are in 2025, with the same prophecies making the rounds.
The reality? 70% of people still trust search engines more than LLMs. Search isn't dying—it's evolving.
But here's what is changing: where and how people search. And that's the key insight that threaded through nearly every session at BrightonSEO.
Search is everywhere (and we need to optimise for everywhere)
Ashley's session on the "search everywhere trifecta" crystallised something we've been feeling at Blu Mint for months now. We call it Brand SEO—the recognition that visibility is no longer enough. The goal is to become the preferred option.
The stats are stark:
76% of consumers use social media to discover brands and experiences
89% of buyers research across at least three platforms before purchase
Searches aren't reducing—they're migrating to platforms like TikTok, Reddit, Pinterest, and LLMs
LLMs aren't replacing Google. They're expanding the discovery process. The blue links era is over, and the multi-platform discovery era is here.
What does this mean practically? At Blu Mint, we're shifting from "how do we rank on Google?" to "how do we develop preference across search, social, and brand touchpoints?"
It's about authority + trust + relevance, achieved through an integrated approach that connects:
Brand-led narratives (the story)
Search foundations (where is this story told?)
Social search and distribution (how do audiences find our story?)
One size doesn't fit all: the harsh truth about AI optimisation
Here's where things get uncomfortable for anyone looking for a simple playbook.
The most important insight from the conference? There is no universal AI search optimisation strategy.
Different large language models—Google AI Overviews, ChatGPT, Perplexity, Claude, Gemini—are trained on different data, use different methodologies, and generate results in fundamentally different ways.
If you're treating them all the same, you're leaving visibility (and results) on the table.
The pre-trained vs search-augmented divide
Generative engines don't rank content—they interpret based on what they've learned. Users aren't clicking links; they're consuming answers. This means we're not optimising for algorithms anymore; we're influencing the architecture that makes content easily digestible.
There are two distinct types of LLMs we need to consider:
Pre-trained models (like Claude, ChatGPT free, most of Gemini) rely on fixed training data sets. They provide base knowledge but are open to hallucination. For these models, your goal is simple: be present in high-trust, longstanding sources when the next training update happens.
Search-augmented models (like Perplexity, ChatGPT paid, Copilot 365) combine pre-trained knowledge with live web data. They're more dynamic, pulling from current search results. For these, traditional SEO still matters—but with a twist.
The retrieval framework: are you present, accessible, and trusted?
The framework that resonated most with me was simple but powerful, much like the approach used by Blu Mint:
Are you present? Audit your presence across platforms. Use tools or do it manually, but know where you appear.
Are you accessible? Is your content structured for AI retrieval? Not for ranking—for retrieval.
Are you trusted? Do you have entity-level trust and authority across high-trust sources?
This means reinforcing entity presence across Wikipedia, Forbes, Reddit, LinkedIn, and Quora. It means restructuring existing content to make it retrievable, not just rankable.
The content that actually works in 2025
Luke Heinecke dropped a bombshell: out of 300,000 articles analysed, only one in three ranked on Google.
But the winning patterns were clear:
Make content easier to extract, not more comprehensive
Fewer words means AI can extract information more efficiently. The days of 3,000-word comprehensive guides dominating every query are over. AI Overviews favour concise, extractable content.
Give people what they actually want
If your page doesn't give people what they came for, it doesn't matter how good it is. People don't want service pages—they want comparison pages. "X vs Y vs Z" content gets 12x more citations than standard blog posts.
Write for humans who know you're using AI (because they do)
AI-written content isn't inherently bad. But if it's written for search engines rather than humans, it fails. The winning approach?
Add specific testing results: "We tested X and found Y"
Include named sources with links
Use real people, real places, and real things (entity-first knowledge)
Answer the intent, prove you're right, then add everything else
Multiple speakers emphasised E-E-A-T signals: Who wrote this? How do they know? Why should we trust them?
Visual content still matters
Above-the-fold infographics resulted in 31% longer time on page. Use AI-generated infographics with client logos. Use multi-model intelligence: ChatGPT for research, Claude for writing, Gemini for designs.
Then get AI to review your content using different models.
The data that matters: publishers dominating AI visibility
Kelsey Libert's research revealed which publishers dominate AI citations—and the results should inform your digital PR strategy immediately.
Five publishers dominate AI visibility overall:
WebMD (~1.2M AI citations)
BBC (~489.6K)
Forbes (~468.2K)
Business Insider (~397.8K)
People (~345.2K)
But here's what's fascinating: model-by-model favourites differ significantly. In ChatGPT, Forbes leads (~202K citations). Perplexity leans into WebMD and Forbes. Copilot tilts to Forbes and MSN. Gemini surfaces WebMD and CNET.
What does GenAI trust most?
The forecast heatmap scores these highest on training value:
Peer-reviewed journals
Patents and standards
University materials
Court and government transcripts
Investigative journalism
Open-source code
At the bottom: SEO blogspam, AI-ghostwritten fiction, fake reviews, clickbait wires, and ephemeral social chatter.
The implications for digital PR are massive. Getting featured in the right publications isn't just about backlinks anymore—it's about training the AI models that will answer questions about your industry for years to come.
Technical foundations still matter (a lot)
Amidst all the AI excitement, Ahmed's session was a crucial reminder that technical SEO fundamentals haven't disappeared—they've become more important.
Crawl shaping matters
30-60% of crawl requests are wasted on low-value URLs. Crawl shaping through XML sitemaps, robots.txt, and internal linking is critical.
Internal linking drives results
Pages with 0-4 internal links get 2 clicks from Google. Pages with 40-44 internal links saw 4x more clicks.
The 80/20 rule applies: the 20% of your pages that bring in the most value should link to priority pages (revenue pages).
Variety always beats repetition when it comes to anchor texts for internal linking.
Content hubs build authority
Organise content around what you're trying to talk about, then build out subtopics. Clear hierarchy matters. Google cares about engagement signals—scrolling, staying on pages, learning through games, tools, graphics, images, tables of contents, and jump links.
The summary? Shape crawls, build internal linking structure, construct content hubs, and engage your users.
The B2B shift: intentful TOFU and the new funnel
For B2B marketers, the message was clear: top-of-funnel (TOFU) content still matters, but the approach needs updating.
Despite AI Overviews dominating informational queries, TOFU content remains essential for brand recognition, resonance, and authority. But impressions are up whilst CTR and clicks have dropped.
The new B2B SEO mindset requires:
Intentful TOFU: niche and long-tail focus
Audit on-page strategy: value-less content is dead
Use subject matter experts: mine intelligence from sales reps, customer service, software developers
Cari's B2B SaaS case study proved this works. After fixing technical foundations (missing titles, meta descriptions, target keywords, internal linking, alt text, and schema), the team asked a crucial question: what new content should we create?
The answer came from directly asking sales, customer service, social media, and marketing teams. This untapped treasure led to a 9% increase in keyword rankings and 25% year-over-year growth in organic traffic.
The content research process:
Investigate using GA4, Google Search Console, Semrush, Screaming Frog
Interview customers: What pain points? What did you wish you knew? Why did you choose this product specifically?
Interview potential customers at trade shows, conventions, social media
Look out for new terms and phrases
Make content easy to understand. Bullet points make it easy for people, AI, and search engines to scan. Implement schema: organisation, blog, services, articles.
Authority architecture: building your engine
Cristiano Winckler's session on crafting your authority engine provided the most comprehensive framework for succeeding in AI search.
The landscape has shifted dramatically:
58% of searches end without a click
26% of users end sessions after seeing an AI summary
LLM traffic will replace organic traffic—but with better conversion rates (8.7% for AI traffic vs 0.8% for low-quality organic traffic)
Three fundamental shifts in SEO for AI search:
Build authority: topical expertise is more important than simple keywords
Content is king: in AI search, rankings are no longer the primary metric
Integration, not isolation: SEO, brand, content, CX, UX, and social media need to be integrated together
The authority architecture framework includes:
Systematic expertise extraction: mine knowledge from sales reps, customer service, software developers
Content creation: build topic clusters with website structure that AI can consume
Entity optimisation: structured presence in the knowledge graph through Reddit, Wikipedia, structured snippets, schema markup
E-E-A-T amplification: build external trust signals through customer reviews, citations, mentions, bios
Integration strategy: unifying authority across every channel
What we're implementing at Blu Mint
Coming back from San Diego, several things are now priorities for us and our clients:
1. Embracing search everywhere optimisation
We're no longer asking "how do we rank on Google?" We're asking "how do we become the preferred option across search, social, LLMs, and emerging platforms?"
2. Building authority engines
We're implementing systematic expertise extraction processes with clients. Your best content insights don't come from keyword research tools—they come from the people who talk to customers every day.
3. Multi-model content strategies
Different LLMs favour different content types and sources. We're tailoring content strategies to perform across pre-trained and search-augmented models, not treating AI search as monolithic.
4. Digital PR focused on AI training sources
Getting featured in Forbes, WebMD, BBC, and other high-authority publishers isn't just about backlinks anymore. It's about being present in the sources that train AI models.
5. Technical foundations with AI retrieval in mind
We're restructuring content for retrieval, not just ranking. That means clear hierarchy, strong internal linking, comprehensive schema implementation, and entity optimisation.
6. Integration as standard practice
The days of siloed marketing are over. SEO, content, social, PR, UX—they all need to work together to build consistent brand authority across every touchpoint.
The uncomfortable truth
Here's what I'm taking away most from BrightonSEO San Diego: the comfortable days of following a standard SEO playbook are gone.
Ross Simmonds put it perfectly: search isn't just happening on Google anymore. People discover content through Reddit threads, TikTok comments, LinkedIn posts, YouTube Shorts, newsletters, niche communities, and increasingly through AI-powered search tools.
CNET made $1.3 million USD from AI traffic alone. That's not a future scenario—it's happening now.
The question isn't whether AI will reshape search. It already has. The question is whether you're building the systems, content, and authority to succeed in this new landscape.
The AI automation opportunity: practical workflows that work now
One of the standout sessions came from Britney Muller, who demonstrated something crucial: whilst everyone's talking about AI in abstract terms, very few are showing you how to actually implement it in your daily workflow.
Her presentation cut through the noise with practical automation workflows using tools that are available today.
Understanding what AI actually means in marketing
First, let's be clear about what most marketers mean when they say "AI":
Ad bidding: reinforcement learning and time-series forecasting
Audience targeting: clustering models
Image/video optimisation: computer vision
Predictive customer lifetime value: regression models
But for content and SEO workflows, we're talking about large language models—and understanding their limitations is as important as understanding their capabilities.
What LLMs are actually good at (and what they're not)
LLMs excel at:
Language translation
Content summarisation
Programming support
Content creation and editing
Classification (like spam detection)
Simplifying complex content
Sentiment analysis
Prompt engineering
Content repurposing (turning blogs into social posts)
SEO meta descriptions and titles
Meeting note organisation
LLMs struggle with:
Real-time or current events
Precise maths and counting
Humour
Consistency
Factual accuracy
Environmental sustainability
Reasoning and logic
Emotional intelligence
Bias-free responses
Extended recall and memory
Understanding these limitations is critical when building automation workflows.
Practical automation workflows you can implement today
Britney demonstrated several workflows that solve real marketing problems:
1. AI-powered news monitoring and content creation
Using Zapier Agents, she showed how to automate the process of staying current with industry trends:
Friday AI News Recap: Triggers every Friday at 8:30am, fetches and analyses headlines from TechCrunch and MIT News from the past seven days, then extracts comprehensive article data including titles, descriptions, and content.
Friday News Dump Analyser: Fetches RSS feeds, parses content, and extracts headlines from the past week across multiple sources like TechCrunch.
2. Turning intelligence into content with GPT for Work
The GPT for Work add-ins bring ChatGPT, Perplexity, Gemini, Claude, Ollama, Grok, DeepSeek, and Mistral directly inside Microsoft Excel and Word, and Google Sheets and Docs for extraordinary productivity.
She demonstrated using Claude's Sonnet 3.7 model specifically for writing tasks, showing how different models excel at different functions.
3. Tools for research and content intelligence
WordCrafter: Analyses what Google rewards for any keyword by examining top-ranking pages, search intent, and content gaps. It also transforms Reddit conversations into deep audience insights and combines all this intelligence with AI controls to craft high-quality content that outperforms competitors.
Ollama for private on-premise LLMs: For brands concerned about data privacy, Ollama allows you to run open-source models locally. One attendee shared they'd built a personal finance app using Ollama, keeping all financial data completely private.
Colab Notebooks: For those wanting to learn Python for AI, Britney offers free workshops and notebooks for hands-on learning.
4. Automating SEO workflows
One of the most practical demonstrations showed how to automatically surface internal link opportunities:
Screaming Frog → OpenAI Embeddings API → Colab Notebook → Google Sheets
This workflow identifies semantically similar content across your site and suggests internal linking opportunities, with the option to add Google Search Console data for better prioritisation using Vlookup.
5. Automated PR opportunity tracking
The PR Opportunity Slack Agent workflow demonstrated how to:
Use a Zapier Agent to parse emails with the label "HARO/PR Ops"
Identify relevant opportunities based on criteria
Reference Blogsmith content and bylines
Push Slack notifications with all the information needed to respond quickly
6. Personalised outreach at scale
The Personalised Outreach Workflow showed how to:
Use Phantom Buster to identify people engaging with competitors' posts on LinkedIn
Feed that data to an LLM through GPT for Sheets
Craft personalised outreach messages via OpenAI
Scale relationship-building without losing the personal touch
The goal: identify people engaging with competitors' posts on LinkedIn and feed data to an LLM to craft personalised outreach messages.
The key insight: keeping pace with industry trends
Britney highlighted a crucial problem: it's hard to keep a pulse on industry trends in the news cycle. Technologies like Flask, React, and SerpApi are constantly evolving, and staying current whilst executing daily marketing tasks is nearly impossible without automation.
Her workflows solve this by automating the monitoring, analysis, and even initial response to industry developments—freeing marketers to focus on strategy and execution rather than information gathering.
Moving forward
Traditional SEO metrics—rankings, clicks, traffic—remain important but incomplete. The new metrics include:
AI citations and mentions
Cross-platform brand searches
Qualified lead conversion from AI traffic
Entity presence across knowledge graphs
Authority signals across trusted sources
At Blu Mint, we're not abandoning what works. We're expanding what's possible.
The agencies and brands that will thrive aren't those with the best AI tools or the most sophisticated prompts. They're the ones that understand this fundamental shift: we're no longer optimising for algorithms. We're building authority that both humans and machines recognise and trust.
And increasingly, we're the ones who know how to build practical automation workflows that free our teams to focus on strategy, creativity, and building genuine authority—rather than drowning in repetitive tasks.
That's what BrightonSEO San Diego taught me. That's what we're implementing. And that's what the future of search looks like.
The revolution isn't coming. It's already here.
