Our BrightonSEO 2025 Takeaways: LLMs, Attribution & SEO Strategy

Last week, a delegation from Ten Speed’s team flew to San Diego to attend BrightonSEO. Marketers from all over the world flocked to California to learn from SEO experts and connect with other marketing professionals across industries. The conference featured some incredible speakers and presentations about the future of search, content strategy, technical SEO, local SEO, website security, and digital PR.

With three tracks running in parallel, our team decided to divide and conquer to capture as much as possible. From the dozens of sessions we attended, a handful stood out - some reshaped how we think about SEO strategy, others challenged assumptions about what the future of search might look like, and several reinforced the foundations every content marketer needs. 

Below are our six favorite takeaways.

Attribution doesn’t need to be a black box, but you do need to take credit for the impacts of content across channels to fuel more on-page decisions.

Perfect attribution is unrealistic because no model or tool can perfectly capture every touchpoint a buyer encounters. But attribution doesn’t have to be binary or directional to be useful. A first step might be defining meaningful benchmarks for success beyond just pipeline and demo bookings (although these are still critical!). 

What does this mean? Content influences discovery, research, and trust well before a customer converts, both on your site and across the web. Measuring those moments gives you a more complete picture of impact and better guidance for on-page decisions.

Content assets, both owned and earned, shape the journey from discovery to engagement, conversion, and pipeline creation. As marketers, we need to account for that full spectrum of influence. For now, we’ll focus on the part we can directly (and indirectly) influence and manage: owned media.

One way to examine this is by working backward through the content journey to spot the “exit ramps” where users drop off and then patching those gaps:

‍

An quick takeaway - LLMs favor content that is less than 12 months old. If this continues to hold true, we should be asking ourselves:

• Are we reviewing everything published across the site in the past 12 months, whether it was shared on social or not?
• What can we learn from how content older than 13 months once performed, and where pipeline may have dropped off?
• Which high-performing posts are now outdated and need a refresh to maintain visibility?
• Where are underperforming pieces dragging us down, and should they be updated, consolidated, or retired?
• Are our most critical customer journey assets staying current and competitive?

Yes. Yes, we should.

Track brand reputation across the web, since visibility drives strength in both LLMs and traditional search.

Earned media often determines whether your brand earns trust or loses it. In fact, 86% of converted users research across three or more platforms, highlighting why consistent engagement is critical.

Brand mentions—both backlinks and unlinked mentions—deserve special attention. They serve as signals of authority and credibility, shaping how search engines and LLMs perceive your brand. Even though 58% of brand visibility opportunities are in zero-click environments (AIOs), you should still strive to stay as visible as possible in the upper funnel. 

For example, before making a major purchase, people often turn to reviews not only on the site itself but also on third-party platforms or via user-generated content on Reddit. It’s a familiar pattern: buyers seek validation from other people they trust before committing. The internet has simply added more layers to the decision process while making the information far easier to access.

It’s not enough for marketers to be plugged in and monitoring these conversations. We need to take the extra step and engage. This could be responding as the brand in reddit, syndicating content, participating in newsletters, responding to reviews, sharing content on socials, and engaging in other online communities. 

Potential customers may hesitate if they encounter inactive social channels or weak engagement in online communities. While you might not control every aspect of earned media, you can still guide the conversation by consistently showing up and participating where it matters or your audience.

LLM traffic is not inherently higher intent than organic traffic, but MOFU content is 

AIOs, AI Mode, and LLMs have gobbled up TOFU clicks over the past 18 months. This started with SGE (search generative experience) in 2023 and continued to build up more momentum as AIOs became more prominent. 

Many keywords in AIOs are low-difficulty (under 40), and 76% already rank in the top 10. Are they worth pursuing? Absolutely. Focusing on keywords in positions 2–10 that aren’t in AIOs can capture greater SERP coverage. When users then see your brand again in LLM results, it helps reinforce trust.

While we are seeing less overall organic traffic, LLMs have not dethroned blue links. In fact, LLM traffic only accounts for 10% of the search market share overall, and less depending on the vertical.

What does this mean? Going all in on LLM traffic as a strategy is premature. The channel is still in flux and will continue to evolve as platforms refine how they respond to changing search behavior.

A few things we do know about LLM traffic today:

• It is more likely, though not guaranteed, to direct users to MOFU assets.
• It is also more likely to surface and cite brands with a strong reputation across the web.

This makes it critical to maintain solid SEO fundamentals and consistent, well-aligned full-journey JTBD content.

LLMs favor fresh content less than 12 months old. 

LLMs show a clear bias toward newer content. Server logs and GA4 data confirm that the most cited and clicked URLs are less than a year old. To strengthen those signals, use a dynamic sitemap and ensure updated posts display “last mod” in the source code.

This doesn’t mean content older than a year will never be cited. But competitor activity is a factor: the more content others in your space are publishing consistently, the greater the risk they’ll be cited instead. In competitive verticals, this makes annual pruning and refreshing even more important.

Another important consideration is how LLMs choose what to surface. ChatGPT, for example, often pulls from Google’s top 20 results. That means whether your content appears in an LLM answer depends heavily on what’s ranking on pages one and two of the SERPs. Sound familiar? Much of what works in GEO builds on SEO best practices.

SEO and GEO follow similar criteria for ranking and referral, but apply them in different ways.

One of BrightonSEO's top insights is that GEO acts on very similar principles with different definitions. GEO evaluates success signals from Google search but weighs them differently. 

Two major shifts stand out when comparing traditional SEO with how LLMs evaluate content:

• Prompts vs. keywords: Traditional search revolves around short keywords of about five words, while LLM searches often rely on prompts that stretch to 20 words or more.
• Citations vs. backlinks: In SEO, almost any backlink could help you rank. In LLMs, the emphasis is on logically placed, highly relevant citations.

The Claude leak confirmed for the industry that TOFU still matters, but it does not drive citations since the model is already trained on that data. What earns citations are MOFU and BOFU: sticky, ICP-centric pieces that generic content cannot replicate. This is because generative engines don’t rank: they interpret.

It’s important to note that LLMs can hallucinate and be manipulated by black hat techniques. To safeguard visibility, you should regularly test the key prompts most relevant to your brand—queries your audience is likely to use—so you can see what the models are citing and adjust your strategy accordingly.

‍

Technical SEO and site speed are no longer nice-to-haves. They’re essential foundations for visibility.

Searching the web has always involved pulling information together from different sources. Since the 1990s, search engines have been the primary intermediary in that process. Today, however, they are no longer the only conduit—LLMs and other generative tools are reshaping how people find and consume information online.

Search engines prioritize sites that are easy to crawl and interpret, and technical SEO is what makes that possible. The advent of LLMs has turned this from a nice-to-have to a sink-or-swim. Sites that are client-side rendered and don’t work with JavaScript disabled cannot be read by LLM crawlers. If this doesn’t send you running to check on your site, it should. 

The next layer is considering how LLMs contextualize site information. They synthesize, so we should set up our content assets to make this process easier. Pages should be built in sections with key takeaways and a table of contents to aid the LLM in grabbing the information the user is seeking based on how it parses the text. Bonus: These elements are also great for UX. It’s a win-win. 

LLMs, like traditional search crawlers, are multi-modal. This means that any content on your site or on owned channels is fair game for a crawl. Videos, images, and audio need the same level of optimization attention as your content since they can all be grabbed and cited in an LLM response. 

Closing Takeaways

A keynote at BrightonSEO captured the connecting thread across sessions: AI is automating manual work, giving marketers the chance to focus on deeper strategy. Tasks that once consumed hours are now measured in minutes, research cycles have been cut dramatically, and the time saved can be directed toward what matters most. 

For marketers, that means getting closer to customers, mapping their journey with greater precision, and aligning content to every step. And this is just the beginning—AI will continue to evolve, reshaping not only how we work but how search itself functions, demanding that our strategies adapt alongside it.