SEO in the LLM Era: Optimizing for AI & Scaling Global SEO

Part 1: Scaling Global SEO in the Age of AI — Key Strategies from DMS & Evolve

Recently at Digital Marketing Summit (DMS) and Ahrefs Evolve, I had the opportunity to lead two sessions focused on cutting-edge strategies in software-as-a-service (SaaS) marketing. One of the sessions zoomed in on advanced global SEO — an increasingly complex landscape shaped by search diversity, AI overviews, and evolving user behavior.

Here’s a recap of the major takeaways from that talk, broken into five core themes:

---

1. The State of Search: What’s Really Changing

Let’s clear up one misconception: Even as Google loses some market share (thanks to players like ChatGPT), the total volume of searches is still rising — +22% year-over-year. That’s the stat that matters.

However, click-through rates (CTR) on informational content are dropping — especially top-of-funnel (ToF) content, which has seen CTRs fall from 1.7% to 0.8%. Why? AI Overviews and search engine result page (SERP) integrations are absorbing clicks. While impressions hold steady, the traffic flow is shifting. Mid and bottom-of-funnel content are less impacted, offering greater return on investment (ROI) for SEO investment.

2. Going Global: The Power of Local Nuance at Scale

Global content strategy should never be one-size-fits-all. I advocate for a global-local sequencing approach:

• Step 1: Localize core content on unique URLs per locale.
• Step 2: Use hreflang tags to connect localized versions and transfer SEO authority.
• Step 3: Build net-new content tailored to specific markets.

Why does hreflang matter so much? It’s not just about showing the right version to the right user. It also passes authority between international pages, influencing rankings — especially in low-authority markets.

Real-world example: Removing U.S.-to-Japan hreflang links caused rankings in Japan to plummet. Fixing them restored performance without touching the Japanese content.

When measuring impact, remember: The average position metric in Google Search Console (GSC) is misleading unless weighted by search volume. 

When averaging keyword positions to assess performance by topic, a simple average gives every keyword equal weight, regardless of search volume. This means a low search volume keyword impacts the average as much as a high search volume one.

𝗘𝘅𝗮𝗺𝗽𝗹𝗲: Consider these keywords with their rankings and search volumes:

𝗞𝗲𝘆𝘄𝗼𝗿𝗱 𝗔: Pos 1, SV 500

𝗞𝗲𝘆𝘄𝗼𝗿𝗱 𝗕: Pos 2, SV 100

𝗞𝗲𝘆𝘄𝗼𝗿𝗱 𝗖: Pos 4, SV 200

𝗞𝗲𝘆𝘄𝗼𝗿𝗱 𝗗: Pos 10, SV 5,000

𝗞𝗲𝘆𝘄𝗼𝗿𝗱 𝗘: Pos 15, SV 10,000

A normal average gives a position of ~𝟲. But a weighted average, factoring in search volume, yields ~𝟭𝟮. That’s a huge difference, and far more accurate, as it prioritizes actual traffic driving terms.

Here’s a simplified version of my custom Excel formula to calculate weighted average rankings:

=SUMPRODUCT(keyword_position, search_volume)/SUM(search_volume)

3. Reclaiming Brand Visibility

If you’re a high-awareness brand, don’t assume you own your narrative online.

News outlets, forums, and even competitors may outrank you for your own brand-related queries. To regain control:

• Create content that addresses those terms (comparison pages, for instance).
• Be aware of local nuances: e.g., in Japan, it’s frowned upon to mention competitors directly (a gentleman’s agreement between brands).

However, one caveat of comparison pages for large language model (LLM) search is that these provide little to no value. LLMs will disregard brand-owned comparison pages as they are specifically looking for neutral content. External comparison pages, however, are still very valuable.

Therefore it is critical to include properly formatted (e.g. table format) and clearly outlined feature and capabilities content on your product page. LLMs (including Gemini/AI Overviews/AI Mode) will be manually comparing your and your competitor’s brand content and making an assessment from there. From ChatGPT below:

4. Page Discoverability: Beyond Just Internal Linking

Don’t focus on the term “internal linking”, it’s too tactical for leadership teams.

Instead, reframe your efforts around page discoverability:

• Create a Page Discoverability Matrix using Screaming Frog or similar.
• Measure how many body-content links (most valuable) each key product page receives.
• Identify directories that receive lots of links but drive minimal traffic or revenue.

Example: A client’s “Resources” section had the most internal URLs but zero links to product pages.

5. Hidden Goldmines: Help & Support Centers

Most SaaS brands neglect their help centers for internal linking — even though these pages often have thousands of backlinks and long-tail traffic.

In my study where I analyzed the top SaaS brands and their site structure:

• Help centers account for just 1.5% of internal links to conversion pages, despite having similar or greater volume than blogs/resources.
  
• Most SaaS Brands could benefit massively by optimizing help content with links to relevant product pages.

One initiative I ran drove a +61% profit uplift, just by optimizing and improving linking in the support center.

👉 PRO TIP: Give your SEO initiatives names — like “Project Cousteau” (a matrix visualizing internal links). Framing it as a strategy makes stakeholders more likely to support it.

---

Part 2: SEO in the LLM Era — How to Optimize for AI Overviews & Chatbots

In my second session at DMS and Evolve, I dived into the world of LLMs and AI-driven search. While organic search still rules; Google handles ~14B searches per day vs. ~34M search-like prompts per day on ChatGPT (for a total of 280M daily prompts, and growing), AI Overviews and chat-based interfaces are already reshaping how brands get discovered.

1. How LLMs Actually Work

Large language models like ChatGPT work by predicting the most probable next word based on context — a concept known as “reasonable continuation.”

To increase your brand’s visibility in these tools:

• Associate your brand closely with relevant terms in high-authority content.
• Citations matter — especially from sources like Wikipedia.
• Think of citations as the new backlinks.

Want proof? Lower the “temperature” in ChatGPT to 0.3, and you’ll see how predictable and brand-safe its responses become, based on its strongest learned associations.

2. Structuring Content for AI Consumption

AI doesn’t care how long your blog post is — it cares about clarity, structure, and credibility.

To be favored in AI Overviews and LLM responses:

• Break content into modular sections.
• Use FAQs, bullet points, and semantic structure.
• Optimize for:
  • Authoritativeness
  • Readability (Flesch-Kincaid scores help)
  • Credible Sources (Quotes, citations)
  • Quantitative Statistics

3. Real SEO Shift: CRO > ToF Traffic

With top-of-funnel CTRs declining, your conversion content must do the heavy lifting.

Focus on:

• Improving conversion rate optimization (CRO).
• Increasing call to action (CTA) clicks instead of just traffic.

When conducting CRO initiatives, focus on the above the fold (ATF) elements, specifically:

• Focus on images with people / product interaction
• Product overviews
• Proof elements (Social, Reviews, Brands, Testimonials) 

4. LLMs Can’t Read JavaScript 

Most LLMs don’t execute JavaScript — except for Google’s Gemini, which uses Googlebot’s full rendering engine. If your content is rendered client-side (React, Angular, Vue), and not visible in raw HTML, LLMs won’t see it.

Fix it by:

• Using isomorphic rendering or server-side rendering (SSR).
• Ensuring critical content appears in the initial HTML payload.

---

Final Thoughts

SEO is evolving — fast.

Your strategies must move beyond just ranking on Google. You now need to optimize for discoverability across AI models, ensure technical visibility for bots, and create content that is both modular and authoritative.

And don’t forget: Conversion > clicks in a world where ToF traffic is thinning out.