How Nexus Protocol
Increased AI Citations by 140%
Users asking ChatGPT "what's the best yield protocol?" were getting recommendations that didn't include Nexus — even though Nexus had better rates and stronger audited security than the names that kept coming up.
Restructured their docs and landing pages around the exact language people use when comparing yield protocols, so LLMs could map Nexus to the right user intent.
+140% Citation Frequency
Verified via GetCiteFlow Monitoring Dashboard
Key Results at a Glance
| Metric | Before | After (6 weeks) | Change |
|---|---|---|---|
| Citation Frequency (ChatGPT) | ~0% of "best DeFi yields" responses | 43% of responses | +Infinity |
| Cross-LLM Citation Rate | Baseline (near zero) | +140% month-over-month | +140% |
| Discord Support Pings | High — "is Nexus any good?" questions | Near zero — users pre-sold by AI | -90%+ |
The problem nobody was talking about
Back in September, Alex (their head of growth) noticed something strange. Nexus was ranked well on Google for most of their target keywords. Organic traffic was fine. But when he pulled up ChatGPT and asked it to compare DeFi yield protocols, Nexus wasn't mentioned — even though their TVL and APY numbers were beating the protocols that did show up.
The issue wasn't SEO. It was that the LLM had been trained on documentation and discussions where Nexus wasn't consistently described using the same language that users type into search prompts. People say "best yield farming protocol" or "highest APY DeFi" — but Nexus's docs talked about "optimized liquidity provisioning" and "capital-efficient pool management." The model couldn't connect the dots.
What we actually did
We started by auditing how LLMs were answering yield-related questions across 40+ prompt variations. For every query that returned a competitor but not Nexus, we traced it back to the source content the model was drawing from. In most cases, the competitor pages used simple, conversational language that matched the query structure almost exactly.
So instead of rewriting their whole site, we focused on three things:
- The comparison pages. We added straightforward comparison tables between Nexus and the top 5 protocols — not just features, but real data points: APY, audit history, TVL, withdrawal fees. The kind of table a user would actually want to see when deciding.
- The FAQ section. We rewrote 12 FAQ entries using the exact phrasing that appears in real user queries. Things like "Which protocol has the highest yield?" and "Is Nexus safer than [competitor]?" — direct, no marketing spin.
- The llms.txt file. We created a structured llms.txt with entity definitions that explicitly connected Nexus to the yield-related terms the model was already familiar with.
The whole thing took about three weeks. A lot of it was just removing the gap between how the product team described their own features internally and how actual users search for them.
What changed
Within six weeks, Nexus was appearing in 43% of ChatGPT responses about "best DeFi yields" — up from basically zero. The citation frequency across all monitored LLMs increased 140% month-over-month. But the real signal was something Alex noticed by accident: their Discord mods stopped getting pinged with "is Nexus any good?" questions from people who'd heard about the protocol through ChatGPT. New users were already pre-sold because the model had done the comparison for them.
"We were spending six figures on SEO and paid acquisition, but the thing that actually moved the needle was making sure the AI could explain what we do in plain English. Once we stopped assuming the model would figure it out and started spelling it out, the results came fast."
— Alex Chen, Head of Growth, Nexus Protocol
Key Takeaways for Your Brand
- SEO does not guarantee AI visibility. Nexus ranked well on Google but was invisible in ChatGPT. The two channels operate on different signals.
- Match your language to user queries, not internal terminology. "Optimized liquidity provisioning" meant nothing to the LLM. "Best yield farming protocol" did.
- Comparison pages with real data are the highest-impact change. Tables with APY, audit history, and fees gave the model structured data it could cite directly.
- Results can come in weeks, not months. Nexus saw measurable citation changes within 6 weeks of making content structure changes.
- AI citations reduce support burden. When the model answers questions accurately, users arrive pre-educated and pre-sold.