Monitor · Answer Engine Insights
See yourself the way the engines see you.
Visibility, prominence, sentiment, citations and share of voice across five engines — computed only from answers Synthic actually received, each stored verbatim and one click away.
Visibility score
58
+12 vs. first audit
Mention rate
53%
240 answers scored
Citation rate
31%
114 links returned
Visibility · weekly
12 wk
Per engine · mention rate
- ChatGPT71%
- Claude58%
- Gemini46%
- Perplexity63%
- Grok29%
Share of voice · who gets named
- Blue Bottle186
- Northwind CoffeeYOU139
- Stumptown97
- Counter Culture58
5
engines, official APIs
5
buyer personas
6
prompts per keyword set
1
click from score to source
Anatomy
How it's built, part by part.
01The chain of custody
Every number has a receipt.
Prompt as sent, answer as returned, verdict on that answer, and the arithmetic that turns verdicts into a score. Any figure on a chart opens to the answer behind it.
02The record
What gets stored, field by field.
The row written for every answer. Nothing is summarised before storage; on paid plans each prompt is asked several times and the agreement is stored as variance, so a lucky answer never becomes a trend.
| engine | text | which engine answered |
| model | text | the exact model id used |
| prompt | text | the question, as sent |
| response_text | text | the answer, as returned |
| mentioned | bool | did it name the brand |
| prominence | 0–100 | how early, how central |
| sentiment | enum | positive · neutral · negative · absent |
| competitors_mentioned | text[] | who else it named |
| citations | url[] | sources it cited |
| mention_basis | enum | retrieval · training |
| runs · mention_count · variance | int · int · real | multi-run agreement |
| latency_ms · capture_method · created_at | int · text · time | the evidence line |
03Personas
The founder and the CFO don't get the same answer.
Phrasing shifts with who is asking, and so do the recommendations. Run one keyword set through five buyer lenses and see where you win, and with whom.
What you get
Three things, done properly.
01
Per-engine visibility
Mention rate, prominence and sentiment for each engine, so you know where you win and where you're absent.
02
Share of voice
Who the engines name instead of you, tallied per prompt and per engine, with every count linked to its answers.
03
Citations & sources
The domains the engines actually pull from — Perplexity's native sources and URLs surfaced in every other answer.
Specification
Inputs, outputs, cadence, evidence.
- Inputs
- Brand domain, up to 3 keywords per set, a buyer persona, the engines connected to the workspace
- Prompts
- 2 per keyword (max 6), built by lib/scoring.ts — inputs to the engines, never answers
- Outputs
- Visibility Score 0–100, mention rate, average prominence, citation rate, per-engine coverage, share of voice
- Evidence
- One audit_results row per answer: engine, model, prompt, verbatim response, verdicts, citations, latency, capture time
- Confidence
- Paid plans ask each prompt 2–5 times; agreement is stored as variance and labelled stable / some / high
- Cadence
- On demand on every plan; weekly scheduled re-audits on Growth and above
- Budget
- Each run has a wall-clock budget; a run that hits it is scored on what completed and marked partial
The answer is being written right now.
Run a live audit and see exactly where you stand — before your competitor does.
- Free first audit
- No credit card
- Live engine data
- Your data stays yours
- Cancel anytime