How does Google flag commercial queries using n-grams in patent US8554638B2?

Has anyone reverse-engineered or tested the way Google assigns weights to n-grams for detecting commercial intent per the 2013 continuation patent? I saw some interesting SERP behavior this month where keywords like 'best headphones under 100' suddenly shifted to more transactional layouts, with product carousels getting boosted. Wondering if that's still evidence of some legacy n-gram scoring or something newer layered on top. Anyone running GSC click data or logs on niche ecomm terms notice similar shifts?

Asked by amir110

2 Answers

Yep, Google definitely used n-grams in that patent to help ID commercial intent, basically by counting term patterns like 'best', 'under', 'buy', etc in queries and assigning weights. I've seen the same thing in SaaS, too, where stuff like 'best X software', 'top X tools', or 'buy X online' tends to trigger more commercial or transactional SERP features, especially product carousels or shopping boxes. But honestly in 2026 the system feels way more layered now, not just a raw n-gram weight system. Machine learning signals seem to play a bigger part, pulling click data and user flow into query intent recalibration. If you look at GSC, try splitting clicks and impressions for comparison queries with and without commercial modifiers over a 6-12 month range. When SERP layouts shift, you'll usually see a traffic bump for commercial queries right after a UI change, especially if you rank in those FAQ/PAAs or get picked up in shopping carousels. I wouldn't be surprised if Google uses a combo of legacy n-gram scoring, but with a lot more real-time user feedback signals on top now. For ecomm, keep tagging and tracking those modifier combos, and watch how Google swaps in different SERP features after core updates.

Answered by bruxellois

Yeah, Google definitely still uses n-gram analysis for commercial intent, but the way it flags and weights those phrases has evolved a ton since the US8554638B2 patent. I see Google stacking newer intent models (like BERT) on top of those legacy n-gram signals. Stuff like 'best headphones under 100' will always trigger commercial SERP shifts because Google recognizes combos like 'best' + [product] + 'under' + [price] as high-buying intent using those n-grams as part of the scoring. In my own client GSC data, click-through and impression spikes on terms with clear purchase language often line up almost exactly with interface changes like more carousels or those FAQ/PAAs pointing to product guides. Not all n-grams are weighted equally anymore though. Phrases that used to signal reviews ('top 10', 'best', etc.) can now trigger more transactional results if there's clear demand. I recommend exporting 6-12 months of Search Console data and segmenting queries by known commercial vs informational n-grams, then check correlation with SERP layout changes. Also try tracking any new merchant or PAA SERP blocks, Google seems to amplify these when enough click signals hit on those n-gram-heavy commercial queries. Realistically, it's a mix of old patent logic and newer ML on top. If you're in ecomm, definitely monitor GSC closely when you see those keyword structure shifts.