Skip to main content
After an AI platform returns a response, Cited has to figure out which brands were mentioned, where they appeared, and how they were characterized. That structured layer is what powers every metric on your dashboard.

​
What the parser captures

For each response on each tracked platform, the parser identifies:
  • Brand mentions. Every brand name in the response, including common variations and abbreviations, brand names embedded in longer phrases, and repeat mentions of the same brand.
  • Position. The order each brand appears in — first brand mentioned is position 1, second is position 2, and so on. This is what feeds average position.
  • Sentiment. How the brand was framed — positive, neutral, negative, or mixed.
  • Framing tags. One to three short descriptors that capture how the brand was characterized (e.g., “reliable”, “expensive”, “design-led”) — useful for narrative analysis later.
A single response can produce multiple brand-level data points (one row per brand mentioned), each with its own position and sentiment.

​
Sentiment, in four categories

Cited reports sentiment as one of four categorical values:
  • Positive — brand described favourably (effective, well-rated, reliable, innovative)
  • Neutral — brand mentioned factually without evaluative language (offers, provides, available in)
  • Negative — brand described unfavourably or criticized (controversial, overpriced, problematic)
  • Mixed — response contains both positive and negative characterization of the same brand
Sentiment is classified per mention, not per response — the same response can be positive about one brand and negative about another. The dashboard rolls these up into a single sentiment score per brand by netting positives against negatives over the time window you’re viewing. We deliberately use four categories rather than a continuous 0–100 sentiment score. The honest read on whether AI just described you positively or negatively is more reliable than a precise-looking number that compresses too much nuance into a single digit.

​
What the parser does NOT do

The parser has deliberate scope boundaries:
  • It does not extract pricing data or feature comparisons — those require category-specific parsing that is not generalisable.
  • It does not equate category-level mentions with branded mentions (“skincare” is not “Plum skincare”).
  • It does not infer brand sentiment for prompts that did not mention your brand. If you weren’t in the response, you contribute to mention rate as a zero, not to sentiment.

​
How Cited references competitors in your AI Narrative

Cited only references competitors you have explicitly declared in your brand profile. When generating your AI Narrative summary, Cited reads the competitor list from your brand settings and constrains the summarization to those competitors. This guardrail exists because, without it, the narrative could surface incidental competitor mentions from raw response content — including outdated, irrelevant, or incorrectly extracted ones. Anchoring on your declared list keeps the narrative aligned with your strategic framing. If you have not declared any competitors, Cited does not name any — it references “the broader category” or “other tools in this space.” Declaring at least 3–5 direct competitors in Settings unlocks richer competitive narrative analysis.

​
Frequently asked questions

Reliable for clearly positive or clearly negative framings, less precise for subtle or mixed framing. The four-category system avoids false precision — when the AI was genuinely ambivalent about your brand, you see “mixed”, not a confident-looking number.
False positives and false negatives both occur at low rates. The parser is tuned for precision over recall — it is better to miss an ambiguous mention than to count a false one. Manual audit spot-checks are part of the quality process.
Yes. The Prompts & Responses view in the dashboard shows the underlying AI responses for each tracked prompt, so you can read the actual text that produced any given mention, position, or sentiment classification.