Results

Analyze how AI models respond to your tracked prompts. View performance metrics, source citations, individual responses, and AI-powered improvement recommendations.

The prompt results page shows you exactly how AI models respond to a specific prompt. Here you’ll see your performance metrics, which sources are being cited, and every individual response.

Header

At the top of the prompt results page, you’ll find key information and actions.

Last updated

See when data was last collected for this prompt. Click Refresh to manually trigger a new run and get fresh results.

Quick actions

Organize and navigate without leaving the page:

• Add tags - Group this prompt with tags like “Competitor Comparison”, “Feature Query”, or “Pricing”
• Attach personas - Assign personas (e.g., “Enterprise Buyer”, “SMB Owner”) to segment your analysis
• View prompt - Jump to the prompt configuration page

Core metrics

The header displays your four key performance indicators:

Citations

How often your pages are cited as sources

Share of Voice

Your percentage of the AI response

Average Position

Your typical ranking when mentioned

Sentiment

How positively AI describes your brand

Each metric shows:

• Current value
• Trend direction (up/down arrow)
• Change from previous period

Improvements

If you have enabled improvements for this prompt, you will see recommended actions to boost your visibility.

The improvements section shows:

Recommendation	Description
Pages to optimize	Specific pages on your site that could help you rank for this prompt
Content gaps	Topics you should cover that competitors are ranking for
Similarity score	How well each page matches the prompt (content relevance, not difficulty)

Understanding similarity scores

The similarity score tells you how well a page content matches the prompt, not how difficult it is to rank.

• High similarity (70%+): Content already relevant. Focus optimization here first.
• Medium similarity (40-70%): Partially relevant. May need expansion.
• Low similarity (below 40%): Different topic. Consider creating new content instead.

Example for “best CRM for sales teams”:

• /solutions/sales-crm at 89% directly addresses the query
• /features/reporting at 45% related but not specific enough
• /blog/marketing-tips at 12% wrong topic entirely

Enable improvements in your prompt settings to get AI-powered recommendations for this prompt.

Sources

The sources section shows which websites AI models cited when answering this prompt.

Source table

Sources are grouped by domain with these columns:

Column	Description
Domain	The source website (e.g., g2.com, wikipedia.org)
Category	Type of source (competitor, review site, publication, etc.)
Share of Voice	What percentage of the answer meaning came from this source
Attribution	How the source was used (direct quote, paraphrase, background)
Position	Where this source ranks among all sources
Citations	How many times this source was cited
URLs	Number of specific pages cited from this domain

Understanding source attribution

Share of Voice for sources shows where the answer meaning comes from, not just which links appeared.

Source	Share of Voice
wikipedia.org	68%
g2.com	18%
your-company.com	8%
techcrunch.com	6%

In this example, even though Wikipedia has just one link, 68% of the actual meaning came from Wikipedia. Authoritative sources often shape the answer even when not prominently displayed.

What to look for

Pattern	What it means	Action
Competitor dominates	Their content shapes answers	Create competitive content
Wikipedia high	Generic info dominates	Add unique, specific details
Review sites prominent	Third-party validation matters	Improve review presence
Your site low	Your content is not used	Optimize pages for AI

Responses

Below sources, you will find every individual AI response for this prompt.

Understanding the table

Each row is one response from one AI model:

Column	Description
Platform	Which AI model (ChatGPT, Claude, Perplexity)
Country	Geographic setting for the query
Persona	Which persona was used, if any
Position	Where you ranked (1st, 2nd, not mentioned)
Mentioned	Whether your brand appeared
Sentiment	How you were described
Date	When captured

One prompt in Attensira creates multiple responses. If you track 2 countries and 3 AI models, that is 6 individual responses per run.

Response detail

Click any row to see the full response:

The detail view shows:

• Full AI response with your brand highlighted
• Sources cited in this specific response
• Competitor mentions and how they were described
• Sentiment breakdown

Filtering responses

Use filters to find patterns:

Filter	Use case
By Platform	Compare ChatGPT vs Claude vs Perplexity
By Country	See geographic differences
By Position	Find where you win (Position = 1) or lose
By Mentioned	Focus on responses where you are missing

Finding patterns

Look across responses to answer:

• Why do I rank better on Claude? Filter by platform, compare attributes
• Why am I missing in the UK? Filter by country, check local competitors
• What makes me win? Filter Position = 1, analyze common factors

Next steps

Winning Prompts

Learn strategies to improve your ranking

Page Results

See performance from the page perspective