LLLLM LEADERS
v1.7---------- · --:--:--ZSCORECARD ONLINE · 4 LIVE · 2 PENDING
CHATGPT · LIVEOK OPENAI · GPT-5o·PERPLEXITY · LIVEOK SONAR-PRO·GEMINI · LIVEOK GEMINI 2.5·CLAUDE · LIVEOK SONNET 4.5·GOOGLE AI · PENDING COMING SOON·COPILOT · PENDING COMING SOON·ANON CAP2/24h PER BROWSER·METHODOLOGYv1.7 DISCLOSED·COMPOSITE SCOREGATED AWAITING FOUNDER SIGN-OFF·CHATGPT · LIVEOK OPENAI · GPT-5o·PERPLEXITY · LIVEOK SONAR-PRO·GEMINI · LIVEOK GEMINI 2.5·CLAUDE · LIVEOK SONNET 4.5·GOOGLE AI · PENDING COMING SOON·COPILOT · PENDING COMING SOON·ANON CAP2/24h PER BROWSER·METHODOLOGYv1.7 DISCLOSED·COMPOSITE SCOREGATED AWAITING FOUNDER SIGN-OFF·
METHOD v1.7⌘K SEARCHRUN SCORECARD ↵
llm-leaders scan--platforms=3-live--prompts=light--depth=anon

Find out how AI

your business — and
fix what they get wrong.

Most firms don't know whether ChatGPT, Gemini or Claude recommend them to buyers. We measure it across three dimensions today, publish every audit, and ship a fix plan. Perplexity, Google AI Overviews and Copilot are coming next — flagged honestly until they ship.

> SCAN
> KEYWORD
No signup requiredLight tier · 2 per browser / 24h4 live platforms · 6 promptsFree
QUERIED AGAINST
GChatGPTMGeminiCClaudePPerplexitysoonAGoogle AI OverviewssoonXCopilotsoon
▸ THIS WEEKFriday breakdown · weekly audit
ISSUE 001 PENDING
No audit published yet

The first Friday breakdown ships when the first consented client audit clears review. We publish every audit with the firm's name on it — never anonymous, never un-checked. Subscribe to be there the day it lands.

Why this is empty: invented case studies discredit the real ones. CLAUDE.md hard rule #1.

F2.1SURFACE COVERAGEper-platform adapter status · crawler v1.7 staging
3 live3 pending
G
ChatGPT
GPT-5o
no data yet
prompts/wk
M
Gemini
Gemini 2.5
no data yet
prompts/wk
C
Claude
Sonnet 4.5
no data yet
prompts/wk
P
Perplexity
Sonar-Pro
Pending
— prompts/wk
status◆ coming
API key pending. Adapter wired; activates when the key is configured.
A
Google AI
AI Overviews
Pending
— prompts/wk
status◆ coming
No stable public API. We refuse to fake coverage via fragile scraping.
X
Copilot
Bing + GPT
Pending
— prompts/wk
status◆ coming
Bing Chat API integration in queue. Shipping when stable.
F2.2CATEGORY WATCHLISTmacro indicators · why this category exists
CLAIMS AUDIT PENDING

The infrastructure most sites assume no longer applies.

We don't editorialise these. They are what they are. The interesting question is what your numbers look like inside this environment.

Every figure below is held back for verification. We will not assert an unverified stat as fact (CLAUDE.md §rule 2). Source-checked values replace these once the claims audit clears.

AI REFERRAL TRAFFIC · H1 250%▲ YoY · Adobe Analytics (reported)
Source candidate: Adobe Analytics, US sample, H1 2025.Unverified · awaiting source audit
GOOGLE SEARCH · FORECAST0%▼ −25 forecast by end of 2026
Source candidate: Gartner research note, Aug 2024.Unverified · awaiting source audit
B2B BUYERS USING AI0%▲ +14pts YoY (reported)
Source candidate: Forrester, Q1 2026, n = 2,142.Unverified · awaiting source audit
F3METHODOLOGYsix dimensions · disclosed in full · model v1.7
144 prompts4 live · 2 pendingv1.7 draft
DIM · 01WEIGHT — TBD

Presence

How often an AI assistant mentions your firm given a category prompt, across 24 prompt variants per platform.

How it scoresWeighting is held at 1.0 across all dimensions until the founder signs off the composite formula. Composite scores are not displayed until then (CLAUDE.md guardrail).
DIM · 02WEIGHT — TBD

Prominence

Position-weighted score. First-named worth more than third-named, worth more than buried in a longer list.

How it scoresWeighting is held at 1.0 across all dimensions until the founder signs off the composite formula. Composite scores are not displayed until then (CLAUDE.md guardrail).
DIM · 03WEIGHT — TBD

Sentiment

How the assistant describes you. Adjectives, qualifiers, hedges — coded against a controlled lexicon. Pending founder sign-off before it counts toward composite.

How it scoresWeighting is held at 1.0 across all dimensions until the founder signs off the composite formula. Composite scores are not displayed until then (CLAUDE.md guardrail).
DIM · 04WEIGHT — TBD

Citations

Which underlying pages the assistant cites. Maps to publishing strategy and source authority.

How it scoresWeighting is held at 1.0 across all dimensions until the founder signs off the composite formula. Composite scores are not displayed until then (CLAUDE.md guardrail).
DIM · 05WEIGHT — TBD

Competitive density

How crowded your category is at the AI layer; which competitors take the recommended slots.

How it scoresWeighting is held at 1.0 across all dimensions until the founder signs off the composite formula. Composite scores are not displayed until then (CLAUDE.md guardrail).
DIM · 06WEIGHT — TBD

Infrastructure

Whether crawlers can read your site: schema, robots, canonicals, render path. If this fails, every other score is structurally capped.

How it scoresWeighting is held at 1.0 across all dimensions until the founder signs off the composite formula. Composite scores are not displayed until then (CLAUDE.md guardrail).
F4SERVICE TIERSfunctionally named · priced visibly · pricing pending founder sign-off
3 tiers£297 floor£1,997 ceiling

Feature

comparison
PRICING →

Audit

one-off
£297once

Audit + Fix + Monitor

monthly · 6-month min
£1,997/mo
SCAN
Scorecard across 4 live AI surfaces
weekly
144 prompt variants
Citation map & source ledger
Competitive density table
Walkthrough call
60 min
weekly
FIX
Schema, robots, canonical fixes
Citation outreach
12/mo
Content briefs
8 topics
Implementation in CMS
MONITOR
Re-score cadence
7 days
Competitor change alerts
Multi-domain tracking
up to 4
Quarterly category report
ACCESS
Direct founder Slack
API access
pending
METHOD ·Sequenced infrastructure → citations → content. We don't write content for sites the AI can't read. If your situation doesn't fit any of these tiers, we will tell you and point you somewhere it does.
Pricing pending founder sign-off — figures above are provisional and may change before launch.
F5CASE REPLAYscroll to scrub · week 00 → 12 · activates with the first consented audit
FIRST AUDIT PENDING
No case replay yet

Case Replay turns on the day our first consented engagement closes. Every audit shown here will name the firm, with their written permission, and the trajectory will be drawn from live scorecards — not reconstructed after the fact.

Why this is empty: a fabricated “11 → 67 in 12 weeks” would discredit the real one. CLAUDE.md hard rule #1.

F6PUBLICATION ARCHIVEevery Friday · on the record · with the client's name on it
ARCHIVE OPENS WITH ISSUE 001
No issues published yet

Every audit ships with the firm's name on it, or with their explicit, written permission to anonymise. The archive fills one row at a time, every Friday, starting with the first consented engagement.

Why this is empty: a fake archive is exactly the kind of thing this firm exists not to do. CLAUDE.md hard rule #1.

Subscribe to the Weekly Breakdown ↗ to receive issue 001 the day it lands.

F7API · COMING LATERscorecards from your stack · out of scope for v1
DESIGN PREVIEW
POST/v1/scorecardRun a new scorecard
GET/v1/scorecard/:idRetrieve results
GET/v1/domains/:id/scoreLatest composite + dimensions
GET/v1/domains/:id/historyScore timeseries
POST/v1/webhooksSubscribe to score events
GET/v1/platformsSurfaces currently tracked
curlnodepythongoruby
The API is out of scope for v1. The endpoint shapes are previewed so retainer customers know what lands when it ships. No beta counter shown until real teams are using it.
POSTapi.llmleaders.ai/v1/scorecardshape · not live
# Shape preview only — the API is not live in v1.
$ curl -X POST https://api.llmleaders.ai/v1/scorecard \
    -H "Authorization: Bearer $LLM_KEY" \
    -d '{
      "url":       "your-business.com",
      "platforms": ["chatgpt","perplexity","gemini","claude"],
      "depth":     "full"
    }'

# Response shape (planned):
{
  "id":        "sc_…",
  "status":    "running",
  "url":       "your-business.com",
  "prompts":   144,
  "platforms": 4,
  "eta_s":     86
}
F7PRINCIPLES — WHAT WE WILL NOT DOlifted from the manifesto · unedited

What we will not do.

A category drowning in hype rewards firms willing to draw lines. These are ours. The list converts more clients than the tier table does.

  1. 01Invent acronyms. "Generative Engine Optimisation" is a category, not a deliverable.
  2. 02Charge enterprise prices for SME work. Or vice versa.
  3. 03Promise rankings on platforms whose rankings we do not control.
  4. 04Publish a case study without the client's name on it — or their explicit, written permission to anonymise.
  5. 05Write content the AI cannot read, then bill you for it.
  6. 06Sell you a retainer your business doesn't need. Most audits don't need one.
  7. 07Pretend the score will keep going up forever. Methodology evolves; sometimes scores reset.
  8. 08Soften this list. If we did, the rest of the site would mean less.
FROM THE FOUNDERa letter to the reader
AY
Ahmad Younis
Founder · London

AI assistants now sit between buyers and businesses in roughly the way Google did between 2003 and 2018. The infrastructural work that won then still mostly wins now — schema, citations, source authority — but the surface is different, the prompts are different, and almost nobody is measuring what's happening with any rigour.

We do. We publish what we find. We charge prices a real business can pay. If we ever start sounding like every other agency in this category, write in and tell us.

hr@llmleaders.ai · London
F8THE WEEKLY BREAKDOWNone audit · one tool review · no fluff · Fridays
OPS · /STATUScrawler v1.7 staging · first audit pending