Internal linking is one of the most underutilized levers in SEO, and it remains just as critical in the age of LLMs. AI-powered search engines like Google's AI Overviews still rely on traditional search infrastructure under the hood: crawling, indexing, and ranking signals like internal links are what feed the models their answers. Most sites treat linking as an afterthought, relying on static navigation or manual link placement. But when done strategically, internal linking can significantly improve how search engines crawl, understand, and rank your pages.
The Linking Agent coordinates five specialized sub-agents (Boost Links, Related Links, Popular Links, Nearby Links, and Top Links), each handling a different linking strategy. We ran a series of controlled A/B tests across enterprise e-commerce and classified sites to measure their real impact. The A/B tests used controlled sets to isolate the effect of each strategy. The results: 8-47% traffic gains depending on the approach. These strategies have since been deployed across over 7.3 million pages.
The Challenge: Links Without Strategy
Most sites have internal linking, but few have strategic internal linking. Common problems include:
- High-value pages buried deep in the site architecture, receiving few inlinks
- Generic "related" links that don't reflect how users actually search
- Static navigation that can't adapt to changing demand or inventory
- No way to measure whether linking changes actually move the needle
"Similar.ai internal linking increased relevance so that we were able to prove that users were able to find what they were looking for more easily, with statistical significance."
— Kabeer Badi Singh, Head of SEO, eBay Kleinanzeigen
The Approach: Data-Driven Linking Strategies
Rather than guessing which pages to link, the Linking Agent uses real data to determine the best targets:
- Google Search Console data to identify pages with ranking potential
- SERP similarity analysis to find topically relevant pages
- Crawl data to understand which pages search engines visit most
- Revenue and conversion data to prioritize high-value pages
Each strategy was deployed as a controlled A/B test, with matched control groups to isolate the impact of the linking changes.
Results by Strategy
Boost Links to Pages Ranking 4-15
The hypothesis: Pages ranking just outside the top positions have proven relevance. A small ranking improvement in this range can dramatically increase click-through rates.
Implementation: The Boost Links agent identified pages ranking in positions 4-15 and doubled their internal links from high-authority pages on the site. Tested on 2 sites across 3 markets.
Results across 64,400 total pages:
| Metric | Improvement |
|---|---|
| Pages crawled | +27% |
| Organic keywords | +12% |
| Impressions | +22% |
| Traffic | +8.6% |
In a second test on another site, we saw even stronger results: 43% more pages crawled, 19% more keywords, and 23% more traffic.
Boost Links to High-Value Pages
The hypothesis: Pages that convert well deserve more visibility. Boosting links to pages with the highest revenue-per-session should grow both traffic and revenue.
Implementation: The Boost Links agent prioritized pages with the highest revenue-per-session, directing internal links from influential pages across the site.
Results across 6,400 pages:
| Metric | Improvement |
|---|---|
| Googlebot crawls | +237% |
| Active pages | +23% |
| Organic keywords | +5% |
| Impressions | +19% |
| Traffic | +47% |
"Lyst used Similar.ai to launch a new internal linking module in a week to let search engine users find the pages they love. In a few weeks, we measured the impact, showing that we had significantly grown the ranking of these pages."
— Simon Dance, Chief Commercial Officer, Lyst
Related Searches and Popular Searches Blocks
The hypothesis: Users and search engines value contextually relevant links. Linking to topically related pages (not just high-demand pages) improves the user experience and ranking signals.
Implementation: The Related Links and Popular Links sub-agents added dynamic linking blocks that use LLMs and SERP data to identify the most relevant destination pages. Deployed across 7.3 million pages.
Key finding: When links were based on topical relevance rather than just demand, pages ranked for more keywords and users engaged more deeply with the site.
"We're working with Similar.ai to leverage machine learning that understands search engine user needs, automatically concentrating internal linking around the pages users love most."
— Dewi Nawasari, Lead Organic Growth Product Manager, Gumtree
Why These Results Matter
The traffic gains from internal linking compound over time. When pages get crawled more frequently, they:
- Index faster when content changes
- Rank for more keywords as search engines understand the page better
- Pass authority to other pages through their own outbound links
Unlike external link building, internal linking is entirely within your control. The Linking Agent can deploy changes across millions of pages in days, not months.
The Methodology
All tests followed a rigorous A/B testing methodology:
- Matched control groups with similar baseline metrics
- Intervention dates tracked to isolate pre/post changes
- Multiple metrics measured including crawls, keywords, impressions, and traffic
- Statistical significance verified before drawing conclusions
This isn't theoretical. These are real results from real sites, measured with real data.
Conclusion
Internal linking works. The data from these A/B tests shows consistent improvements across different sites, markets, and linking strategies:
- 8-47% traffic gains depending on the approach
- 237% more Googlebot crawls for high-value page boosting
- Statistically significant results across all A/B tests
- Now deployed across 7.3 million pages across enterprise sites
The key is treating internal linking as a strategic lever, not an afterthought. The Linking Agent and its sub-agents ensure that your best pages get the visibility they deserve, automatically.
Want to see what the Linking Agent could do for your site? Request a demo or try our growth calculator to estimate your revenue opportunity.