Sample Audit

This is what an AI Visibility Audit looks like.

We audit ourselves by the same standard we apply to clients. Below is Uygen's own June 2026 baseline — real findings, real scores, real roadmap. The Weekly Tracker tab shows a sample citation-tracking report from a real client audit (anonymized).

To understand the measurement behind this report, read the AI search visibility tracking playbook and the Perplexity SEO checker workflow.

Self-audit · uygen.com Scored 2026-06-15 Next re-score 2026-07-15 Scoring: AVS framework — Access / Understanding / Authority

AI Visibility Score (AVS)

40 / 100

Emerging Band (26–50)

Cited occasionally or not yet; one weak channel. The priority is off-site presence and completing the measurement baseline. Widen third-party presence and run the full citation-pool prompt set before re-scoring.

Raw score: 44 / 110 max — weighted formula: (5×1.5 + 5×1.5 + 15×1.0 + 10×1.0 + 8×0.5) ÷ 110 × 100

Layer	Score	Weight	Weighted	Notes
Citation Share	5 / 20	×1.5	7.5	No tracker cycle complete; baseline estimated for new brand
Third-Party Presence	5 / 20	×1.5	7.5	No independent citations; ~2-month-old brand at audit date
Accuracy	15 / 20	×1	15	Content accurate and unambiguous; unverified via live AI queries
On-Site Extractability	10 / 20	×1	10	Answer-first blog structure; strong data inventory; schema unconfirmed
Measurement Maturity	8 / 20	×0.5	4	Tracker infrastructure built; not yet running at scale
Raw total / AVS			44 → 40.0	Max raw: 110 · AVS = raw / 110 × 100

Findings

P1 = blocking · P2 = pass / corrected · P3–P5 = high to medium priority

Cloudflare auto-injecting Disallow for GPTBot, ClaudeBot, Google-Extended

Layer: Access · Effort: Low

The source robots.txt is correct — but Cloudflare's managed content block injects Disallow: / for GPTBot, ClaudeBot, and Google-Extended above the manual Allow rules. RFC 9309 first-match-wins parsers will treat these bots as blocked. Live citation crawlers (Bingbot, OAI-SearchBot, PerplexityBot) are unaffected and clean. Fix: Cloudflare dashboard → Security → Bots → AI Scrapers and Crawlers → set the three bots to Allow.

llms.txt confirmed live — initial fetch returned a false negative

Layer: Access

/llms.txt exists and is well-structured. Covers key pages, the audit framework, and contact info. A follow-up fetch confirmed the file is live. No action needed.

Citation share unmeasured — tracker running 3 prompts instead of 30

Layer: Measurement · Effort: Medium

The tracker seed contains 3 informational prompts. A citation-share baseline requires 30 buyer-intent prompts across 5 intent categories: brand direct, category awareness, problem-aware buyer, solution-seeking, and comparison. Without this, Citation Share cannot be scored accurately.

Third-party presence thin — no reviews, Reddit threads, or press coverage

Layer: Authority · Effort: Medium

No testimonials on the audit page. No G2, Trustpilot, or equivalent reviews. No Reddit threads identified. Competitor list in tracker seed lists BrightEdge and Conductor (enterprise SEO platforms) — not actual AI visibility peers. Brand is ~2 months old at audit date — this is a launch-phase gap, not neglect.

Schema markup unconfirmed on key pages

Layer: Extractability · Effort: Low

JSON-LD schema is implied by page structure (FAQ blocks, blog posts with structured data) but could not be confirmed from page fetches alone. FAQPage schema on the audit page and BlogPosting schema on blog posts must be validated via Google's Rich Results Test.

Priority Roadmap

#	Action	Layer	Effort
1	Fix Cloudflare AI Scrapers setting — allow GPTBot, ClaudeBot, Google-Extended	Access	Low
2	Seed tracker with 30 prompts from citation pool; run first weekly cycle	Measurement	Medium
3	Validate JSON-LD schema on homepage, audit page, and blog posts via Rich Results Test	Extractability	Low
4	Fix competitor list in tracker seed to actual AI visibility peers (AthenaHQ, Profound, EarlySEO, Otterly, HeyEmmett)	Measurement	Low
5	Add 3+ testimonials to /ai-visibility-audit/ page	Authority	Medium
6	Map Reddit + review presence; identify citation pool subreddits	Authority	Medium
7	Re-score AVS after first tracker cycle; establish monthly cadence	All	Ongoing

Score Trajectory Targets

Period	Target AVS	Band	Key lever
Jun 2026	40	Emerging	Baseline
Jul 2026	45–50	Emerging	robots.txt fixed; tracker at 30 prompts; llms.txt confirmed
Aug 2026	52–58	Developing	2+ Reddit threads; schema validated; Bing WMT confirmed
Sep 2026	60–65	Developing	3+ testimonials; 1 editorial mention; citation share at 10–15%
Nov 2026	68–72	Established	Consistent citation across 2 engines; growing third-party presence

Real client · anonymized Week of 2026-06-16 30 prompts × 5 engines = 150 runs B2B manufacturer · 5 competitors tracked

One-paragraph summary: 30 real purchase-intent queries were run across 5 AI engines (150 total). The client's domain was cited 47 times — but ~32 of those came from Perplexity alone. ChatGPT cited the domain 1/30 times; Gemini 0/30. The gap is not product quality — it's three fixable inputs: no answer-type content (buying guides, FAQ, case studies), inconsistent entity identity across platforms, and structured data missing site-wide.

Citation by AI Engine

Each engine ran the same 30 queries

AI Engine	Cited (gave link)	Mentioned	Read
Perplexity	15 / 30	13 / 30	Only stable citing engine — live retrieval type
Claude	4 / 30	4 / 30	Identity and comparison queries only
Google AI	2 / 30	4 / 30	Category queries dominated by competitors
ChatGPT	1 / 30	5 / 30	Recognizes brand; rarely provides link
Gemini	0 / 30	5 / 30	Mentions name; never cites the domain

Core finding: citation visibility is concentrated in Perplexity (live-retrieval engine). In model-weight-dominant engines (ChatGPT, Gemini), citation requires answer-type content, not just a domain listing.

Share of Voice

Citation counts across all tracked brands — same 30-query set

Brand	Citations	SOV
Client (you)	47	≈ 20%
Competitor A	120	≈ 50%
Competitor B	31	≈ 13%
Competitor C	18	≈ 8%
Competitor D	14	≈ 6%

The top competitor holds ~50% of citations in this query set — 2.5× the client's share. Third-party platforms (B2B directories, YouTube) appear frequently alongside brand domains and are included in the full report citation pool.

Queries Where Client Was Cited

These are the positions to defend first

Query (anonymized)	Engines that cited client
Who is [Client] and what does [client].com sell?	Claude · Google AI · Perplexity
[Client] vs [Competitor A] — which is better for [category]?	Claude · Perplexity
Is [Client] a reliable [category] manufacturer?	Perplexity
Who is a reliable [category] manufacturer in China?	ChatGPT · Perplexity
Which [category] factory offers OEM with low / no MOQ?	Perplexity

Brand-direct and comparison queries are the strongest footholds. Schema markup and FAQ content on the matching pages lock these positions in.

Root cause 1

No answer-type content

Competitors win category queries with buying guides, FAQ pages, and brand listicles. The client has product pages only. AI cites answers, not listings.

Root cause 2

Unreadable page structure

Zero JSON-LD schema site-wide. Product specs in images, not HTML. Title tags stuffed with keywords rather than written for the buyer's query.

Root cause 3

Entity identity fragmented

Three different addresses registered across platforms. Two separate brand profiles on B2B directories. AI systems weight entity consistency as a trust signal.

Want this for your brand?

Book an AI Visibility Audit to find out whether the real blocker is access, extractability, or authority — and get a prioritized roadmap rather than guesswork.

Book a call View methodology