Logging Citation Position and Sentiment, Not Just Presence
Being named first as the recommendation and mentioned last with a caveat both log as cited. How to track position and sentiment, and why each needs a different fix.
Logging only whether you were cited throws away most of the useful information in an AI answer. Two citations can look identical in a spreadsheet and mean opposite things: being named first as the recommended option is not the same as being listed fourth among alternatives, and “a solid choice for enterprise teams” is not the same as “cheaper but more limited than the others.” Recording position and sentiment alongside presence turns a binary tally into something you can actually act on.
Key takeaway
- Presence alone hides whether you were the recommendation or an also-ran in the same answer.
- Log position (where you appear in the answer) and sentiment (how you’re described) alongside citation.
- Sentiment problems and position problems have different causes and different fixes, so tracking them separately matters.
Why presence is too blunt
A binary “cited: yes” flattens genuinely different outcomes into one value. Being the first option an engine names, described as the strongest fit for the situation in the prompt, is close to the best result available. Being mentioned last in a list of six, with a caveat attached, is technically the same data point and a far weaker commercial position. If your reporting can’t distinguish them, you can’t tell whether a rising citation rate reflects real progress or just more appearances in the tail of answers.
Three fields, not one
Cited, position, sentiment. Adding two columns to your tracking sheet converts a flat count into a picture of how favourably you’re being presented — and shows which of the two needs work.
Source — AI visibility measurement practice
Logging position
Keep it simple enough to record consistently across dozens of prompts. A three-value scale works well: primary (named first, or presented as the recommendation), listed (included among several options without particular emphasis), and passing (mentioned briefly, in a caveat, or as an aside). Record which one applies for each run. Over time the distribution matters more than any single result — a brand moving from mostly passing to mostly primary on a topic cluster is making real progress, even if the raw citation count barely moves.
Logging sentiment
Sentiment is how the engine characterises you when it names you. Positive means the description supports choosing you: strong fit, well-regarded, good for this use case. Neutral means factual description without evaluation. Negative means the mention actively works against you: limited, expensive relative to alternatives, better suited to someone else. Record the category and, critically, paste the actual phrasing. The exact wording is where the diagnosis lives — “good for large teams” when you serve small businesses is a positioning problem no sentiment label alone would surface.
Different problems, different fixes
Separating the two dimensions tells you what to work on. Consistently poor position with neutral sentiment usually means a content and corroboration problem: you’re known but not compelling, so the fix is stronger evidence, better third-party presence and content that addresses the evaluation criteria directly. Negative or inaccurate sentiment points at something else — outdated information on the web, unclear positioning, or a genuine weakness competitors are being credited with handling better. And a description that’s simply wrong is a factual correction task, traced back to whatever stale source produced it.
Keep the record usable
Add three columns to the sheet you already use for share of citation: position, sentiment, and the verbatim phrasing. Because AI answers vary between runs, log each run separately rather than averaging on the fly, then summarise per prompt at the end of the round. Reviewing the collected phrasing every month is often more informative than the numbers themselves — patterns in how engines describe you show up in the language long before they show up in the citation count.
Frequently asked questions
Why track citation position and sentiment?
Because presence alone flattens very different outcomes into one value. Being named first as the recommendation and being mentioned last with a caveat both register as “cited,” despite being commercially opposite. Logging position and sentiment shows whether a rising citation rate reflects genuine progress or just more appearances in the tail of answers.
How should I record position?
Use a simple three-value scale you can apply consistently: primary (named first or presented as the recommendation), listed (one of several options without particular emphasis), and passing (a brief mention, aside or caveat). Record it per run. The distribution over time matters more than any single result.
What counts as negative sentiment?
Any characterisation that works against choosing you — described as limited, expensive relative to alternatives, or better suited to a different buyer. Always paste the verbatim phrasing alongside the label, because the exact wording carries the diagnosis. “Good for large teams” when you serve small businesses is a positioning problem a sentiment category alone would hide.
Do position and sentiment problems need different fixes?
Yes. Weak position with neutral sentiment usually signals a content and corroboration gap — you need stronger evidence, better third-party presence and content addressing evaluation criteria. Negative or inaccurate sentiment points instead at stale information, unclear positioning, or a genuine weakness, and wrong descriptions are a factual correction task traced to their source.
The bottom line
Add two columns and a phrasing field to your citation tracking. Position tells you whether you’re the recommendation or an afterthought; sentiment tells you whether being named is helping or hurting. They have different causes and different remedies, and the verbatim wording you collect along the way is usually the most diagnostic thing in the whole report.
We track how prominently and how favourably AI engines describe you, not just whether they do. Part of our AI Visibility service.
