Five sources, one answer

A citation that supplied the AI’s recommendation is not equal to a citation that supplied a background statistic. Most benchmarks count them the same.

When an AI responds to a buyer’s comparison question, it typically cites three to seven sources. The standard treatment is a flat list: count how many are yours, how many are your competitor’s, compute a ratio, move on. That’s source balance, and it’s useful the way a batting average is useful: directionally correct, deeply incomplete.

One source might have supplied the recommendation. Another might have provided a single statistic. A third might appear in the footnotes without visibly influencing anything the buyer reads. Counting them as equivalent is like weighting a walk the same as a grand slam.

The question isn’t how many sources were yours. It’s which sources shaped what the buyer heard.

What is source gravity?

Source gravity is how much influence a cited source had over the substance of the AI’s response.

A high-gravity source is one the AI paraphrased extensively, drew its recommendation from, or used to frame the competitive comparison. A low-gravity source is one that appeared as a footnote, provided a single data point, or was listed for completeness without shaping the narrative.

The distinction matters because the owner of the highest-gravity source effectively controlled the answer.

Consider a response where the AI recommends your competitor for enterprise compliance. It cites five sources: your product docs (background on your feature set), two of your competitor’s blog posts (the basis for the recommendation and the feature comparison), an analyst report (a supporting claim about market positioning), and a community thread (a user opinion that reinforced the recommendation).

Source balance says 1:2:1:1. Looks like a normal distribution.

But the gravity distribution tells a different story. The competitor’s blog posts shaped the recommendation and the feature framing. Your docs provided background that the AI acknowledged but didn’t act on. The analyst report and community thread reinforced a conclusion that was already formed.

The competitor owned the high-gravity sources. You owned a low-gravity one. The answer was written from their frame.

The ownership-influence matrix

This creates a 2D problem, similar to the accuracy-conviction matrix in our scoring system. One axis is ownership (yours, competitor’s, third-party). The other is influence (how much the source shaped the answer).

The top-left and bottom-left cells are where the action is. When a competitor’s high-influence source drives a recommendation, the fix is specific: write the page that answers that exact buyer question from your frame, structured for AI extraction, not for human browsing. When a third-party high-influence source carries the answer, you need to either influence that source or create first-party content that competes with it.

Flat source balance misses all of this. A 3:2 ratio in your favor sounds good until you realize the competitor’s two sources shaped the recommendation and your three provided background no one acted on.

Where this goes

Today the ownership axis is fully built. Every cited source is classified as yours, your competitor’s, or third-party, per workspace, with subdomain matching, and when you claim or reassign a domain, existing citations are re-classified. The Sources view shows the ownership balance across all evaluations, each competitor has its own cited-source list, and you can see which domains each engine actually cites for your category. That’s the foundation. The influence axis is where we’re headed.

Influence-weighted source balance. Instead of counting citations, weight them by how much each source contributed to the recommendation, the feature comparison, and the competitive framing. A competitor-owned source that shaped the recommendation counts more than three of your docs pages that provided background facts.

Source authority scoring. Some sources get cited repeatedly across evaluations. If the same competitor blog post shows up in 15 different AI responses about your category, that page has outsized authority in the model’s understanding of your market. Surfacing these high-authority sources tells you exactly which pages to compete with.

Content gap prioritization. When you know which buyer questions are being answered by competitor-owned high-influence sources, you know which pages to write first. Not the pages with the most missing keywords. The pages that would shift the highest-gravity citations from their column to yours.

Source balance tells you who got cited. Source gravity tells you who controlled the answer. We think you need both.