Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.onsomble.ai/llms.txt

Use this file to discover all available pages before exploring further.

The Individual responses view is the raw material behind every scan. While the other views — citation benchmarks, share of voice, sentiment — summarise what’s happening across the scan, this view lets you read the actual words each AI model produced. Sitting with individual responses is where the most useful insights often come from. Summaries tell you that something is off; the responses tell you why.

When to use this view

Reach for individual responses when:
  • A metric surprised you and you want to see what’s behind it
  • A prompt looks like it should perform well, but the business is absent or poorly represented
  • You’re preparing a client-facing report and want specific, quotable examples
  • You want to understand how competitors are being framed — not just that they’re winning, but how
The view lets you browse responses across two dimensions:
  • By prompt — see how every model answered a specific question. Good for understanding a single customer moment across the competitive landscape.
  • By model — see how one model answered every prompt. Good for spotting model-specific patterns.
Filters narrow the view down — for example, show only responses where the business isn’t mentioned, or only responses with negative sentiment. Use these to get to the interesting cases without scrolling.

What to look for

Reading an AI response with an analytical eye is a skill. A few things that consistently pay off:

Absence

If the business isn’t mentioned at all in a response, read what is there. Which competitors are named? How are they described? What are they being credited with? The gap between “mentioned” and “not mentioned” is the single highest-leverage insight a scan produces.

Framing of competitors

Pay attention to how competitors are described — not just that they’re named. A competitor consistently described as “the established leader in X” is being positioned differently than one described as “also available.” The framing tells you what signals the AI is picking up on and where those signals came from.

Inaccuracies about the business

AI models frequently get specifics wrong — outdated pricing, incorrect product lineup, wrong service area. When they do, it usually traces back to out-of-date or ambiguous source material on the web. Flag inaccuracies as you find them. Correcting the underlying sources is one of the most reliable ways to improve how you’re represented.

Omissions

Sometimes the business is mentioned, but the mention is missing the thing that actually differentiates it. A specialist firm described as a generalist. A premium provider described as a budget option. These aren’t errors exactly — they’re gaps in the signal the AI is working from.

Source attribution

Some responses cite sources explicitly (Perplexity does this consistently; others do so occasionally). When sources are cited, check them. If the cited pages are old press releases, a competitor’s website, or third-party reviews, that tells you where the AI’s picture of the business is actually coming from.

Turning observations into recommendations

As you review, keep a running list of what you noticed. Patterns usually emerge quickly:
  • “Our pricing page is out of date — three responses referenced prices that haven’t been accurate for a year.”
  • “We’re being described as a generalist, but our website positions us as specialists. The specialist positioning isn’t making it through.”
  • “Competitor X is winning on ‘best for enterprise’ prompts — their case study library is visibly richer than ours.”
Each of those is a candidate action. The Acting on recommendations page explains how to turn this list into a prioritised plan.
When reviewing for a client, save or export the specific responses that illustrate your recommendations. Showing the client exactly what an AI assistant said — in quotes — is far more persuasive than reporting numbers in the abstract.