01 · Levers you control
The inputs that decide how AI sees you
- ·Brand foundations
- ·Owned editorial
- ·Third-party signal
- ·Topic ownership
- ·Source mix
- ·Competitive presence
The product
Measurement you can audit. A plan you can act on.
AuthorityOn.ai runs real prompts against the major LLMs every week, scores your brand on the Signal Five, and turns the result into a prioritised list of fixes. It scopes to your market. Global for enterprise brands. Country or city for SMEs and local-services firms. You see the competitors AI actually puts you next to. Here is how each piece works.
How the methodology works
Levers you control. Measured. Turned into a plan.
Most AI-visibility tools start with symptoms. We start with the inputs you can actually move, six levers that decide how models see you, measure them as the Signal Five, and hand back a prioritised plan. A causal chain, not a leaderboard.
02 · The measurement
Signal Five
- ·Spontaneous recommendation
- ·Head-to-head win rate
- ·Discovery & indexability
- ·Source attribution
- ·Recommendation drift
03 · The output
A prioritised action plan
The next ten things to fix, each tied back to a lever above and the signal it moves, ordered by impact. Refreshed every week as the measurement updates.
What you get
Every view into how AI sees your brand. One action plan.
Nine views carry the weekly loop, plus the bonus surfaces your team will live in. Each view is a working dashboard, not a one-off PDF. Updated weekly from real prompts run against every frontier model.
One number, scanned every week.
A composite 0–100 score across every frontier model, refreshed weekly. The single line on the dashboard your CMO will ask you about.
- 0–100 composite, methodology auditable
- 14-, 30- and 90-day trend views
- Drift alerts the moment it moves
Composite score · last 14 days
Composite AI Performance score over time
Every frontier model, scored separately.
Some models already know you, others don't. You see which is which, so you can focus content effort on the models that don't yet recall your brand.
- All 8 frontier models on the same 0–100 scale
- Variant-level (Sonnet vs Haiku, Pro vs Flash)
- Drill into any model for its raw responses
Model coverage
- Perplexity95
- Gemini89
- Mistral86
- Claude82
- ChatGPT81
- Copilot81
- DeepSeek79
- Grok67
Anchored to your buyer's market.
Every prompt is interpolated with your brand's city and country, so the model answers the question your local buyer would actually ask, not the global default. A Geneva firm and a London firm get two different scoreboards.
- Country + city anchoring on every prompt
- Unique to AuthorityOn. Competitors are country-only at best
- Locality history surfaces when the buyer pool shifts
Locality-aware prompt
PROMPT TEMPLATE
“What are the best private banks for high-net-worth clients {{ in Geneva, Switzerland }}?”
Claude✓ Mentions Geneva-anchored players
Perplexity✓ Local + global mix
The stories AI keeps telling.
Every model response is clustered into coherent stories: positive, neutral, critical. You read the shape of the answer, not just the score, plus exactly which claims are moving in your favour or against you.
- Positive / neutral / critical clusters per scan
- Source citations behind every claim
- Mute, resolve, watch the critical ones drift over time
Stories about your brand
Clustered from 200+ model responses this scan
Your brand, plotted alongside everyone else.
Competitors, products, leaders and campaigns rendered as a single map. Sphere size is the AI score, edges are co-mentions. Click any node to see how AI describes that entity, at the same level of detail as your own brand.
- Up to 12 tracked entities on Growth, more on Scale
- Sub-entities scored independently of the parent brand
- Co-mention edges show the context each entity appears in
Entities map
Brand + tracked entities, AI score = sphere size
The next ten things to fix. Sized by lift.
Every audit ends with a prioritised inbox: each action tagged by strategy (content / authority / discovery / earned media), each with an estimated lift on the composite score. Triage like email. Done, snooze, mute.
- Prioritised by projected composite lift
- Strategy + effort tags per action
- Per-brand suppression so wrong recs never come back
Recommendations inbox
Next ten actions, sized by composite-score lift
Publish FAQ schema on /pricing
+12compositeAdd ClaudeBot to robots.txt allow-list
+8compositePitch 3 Tier-1 analyst firms on the new product line
+6composite
Drift alerts the moment the score moves.
Models retrain. Sources fade. Competitors ship. Set thresholds per metric per entity and we fire to Slack and email on the first breach, with the diff explained so you know exactly what to look at.
- Slack + email channels per workspace
- Per-metric, per-entity threshold control
- The diff is in the alert. Never just a 'something changed'
Drift alert
#brand-authority·AuthorityOn.ai
Recommendation rate dropped 14% on ChatGPT
Likely cause: a new competitor citation on Forbes / TechCrunch in the last 72h is showing up in ChatGPT's recall pool for your category prompts. Open the diff for the full trace.
Audits beneath the score.
Three companion audits explain WHY the composite moves: web crawler access + schema, content citability + freshness, social presence + channel mix. Each scored 0–100 with its own action list, rolling up into the headline number.
- Web · Content · Social, each independently scored
- Every audit has its own prioritised fix list
- AI-crawler discovery surfaces blocked bots
Audits beneath the score
Three companion audits explain WHY the AI score moves
- WEB AUDIT72/ 100Strong
Crawler access, schema, render parity, page-speed.
- CONTENT AUDIT58/ 100Mixed
Citability score, AI-extractable structure, freshness.
- SOCIAL AUDIT81/ 100Strong
Cross-channel presence, voice consistency, citation share.
Whether AI visibility drives real traffic.
Connect GA4, Plausible, Matomo or Piwik PRO and we isolate the sessions arriving from ChatGPT, Perplexity, Gemini and the rest, then overlay that AI-referred traffic on your AI Score — so you can see whether rising visibility is moving real visits, and read an estimated traffic lift on each recommendation. Growth and up.
- AI-referred sessions, isolated by source
- Overlaid on your weekly AI Score to show the correlation
- GA4, Plausible, Matomo or Piwik PRO
AI-driven traffic
AI Score ↔ AI-referred sessions: r = 0.74
Sessions arriving from ChatGPT, Perplexity and Gemini, isolated from the rest of your traffic and overlaid on your weekly AI Score, so you can see whether rising visibility is moving real visits.
Plus
Every other capability your team will live in.
Topic leaderboards
Track real buyer questions in your category: "best AI brand-monitoring tool", "top ESG consultancies". See your rank, competitors' ranks, the citation sources behind both.
Custom prompts + workbench
Growth and upTest a prompt against every model on demand, see verbatim responses, then promote it to a Prompt Group so it joins the weekly scan.
Shareable report links
Generate a token-gated public URL for any report, scoped to expire on a date you set. Share with your board or agency. No platform account needed at the other end.
The Pulse, our daily editorial
A free daily briefing on AI model launches, citation-pattern shifts, brand-visibility moves. We watch the AI search shift so you don't have to.
EU & Swiss data residency
EU residency by default; Swiss residency available on Scale and Agency. SOC 2 controls, GDPR + nFADP compliant.
Complete coverage
Every model your buyers are actually using.
We query up to eight major LLMs every week with a curated set of discovery, comparison and recommendation prompts. Real prompts. Real responses. Kept on file.
8
Models covered
200+
Prompts per scan
Weekly
Refresh cadence
ChatGPTClaudeGeminiGemini ProGrok 4MistralPerplexityDeepSeekCopilot
How an audit runs
Three steps to a prioritised action list.
Most GEO tools stop at a leaderboard. We hand you the next ten things to do, in order.
- 01
Connect your domain and your competitors.
Drop in your URL and the brands you benchmark against. We pick the prompt set automatically from your category. No tag, no SDK, no DNS change.
- 02
We run the full audit.
We query every major model with the prompt set, audit your crawler access, score citation sources, and benchmark against your competitors. Real prompts, real responses, kept on file.
- 03
Get a prioritised action plan.
One clear list: what to fix on the site, what to publish, where to earn coverage, what to monitor next. Sized by lift, ordered by what moves the needle first.
Sample report · acme-standards.org
LIVE48/ 100
Composite Signal Five score
Spontaneous recommendation
12%
Head-to-head win rate
84%
Discovery & indexability
60%
Source attribution
41%
Recommendation drift
n/a
See it run on your brand.
Starter is €49/mo, one brand, the full Signal Five, your first action plan ready the moment you sign in. Cancel any month, no notice.