Public demoNo API keys required.Trust & limits

CiteSight

Harbor Loaf

harborloaf.exampleartisan bakery
BusinessAppear
68

Overall · Solid

Demo runAppear

Diagnosis

Missing from category answers

Ask who leads artisan bakery and Harbor Loaf does not show up.

What AI says / gets wrong

  • Claude: Answer does not clearly reflect the claimed category “artisan bakery”.
  • Perplexity: Answer does not clearly reflect the claimed category “artisan bakery”.
  • ChatGPT: Answer does not clearly reflect the claimed category “artisan bakery”.

+2 more after you start a fix

Authoritative sources

Known public pages CiteSight fetched for the record — not buried in probe cards. If AI answers contradict these, treat them as the wrong identity.

Homepage crawl failed (fetch failed). Wikipedia, LinkedIn, or other public pages above are the fallback record.

Act now5 fixesNext clicks for Harbor Loaf

Next step

Watch this name every week

Monitor re-runs these probes, keeps score history, and flags description flips. $49/mo. Two clicks from this report.

Demo checkout — no card required
Work plan

Act now

Top fixes for Appear / grow — one click each. Expand a card only if you need the how-to.

0 in plan · 9 todo · 0 in progress · 0 done

  1. 01

    Earn “best of” listicle, roundup, and press mentions

    critical · P0Effort L

    Pitch or create newsworthy reasons to appear in category roundups, notable-people lists, and cause roundups that AI search engines already cite.

  2. 02

    Lock a single public name

    critical · P0Effort M

    Use one official name everywhere — site, directories, knowledge panels, speaker bios, and press kits. Add “also known as” only when it is real.

  3. 03

    Seed a trustworthy review footprint

    critical · P0Effort M

    Collect verified reviews on the platforms people (and models) already read, and respond to negatives in public.

  4. 04

    Ship honest comparison and alternatives pages

    critical · P0Effort M

    Create /compare/{peer} and /alternatives pages that models can quote when users ask “X vs Y” or “alternatives to X”.

  5. 05

    Add schema.org Organization JSON-LD

    high · P1Effort S

    Emit Organization markup with legal name, URL, logo, sameAs profiles, and founding data so crawlers and answer engines can ground the entity.

All 9 tasks
Seed a trustworthy review footprint
Collect verified reviews on the platforms people (and models) already read, and respond to negatives in public.

Why it matters

“Is {name} trustworthy?” answers track review sites, news, and complaint boards more than a homepage.

How

Ask recent customers for G2/Google/industry reviews. Publish security/compliance pages. Do not gate or incentivize dishonestly.

Details

Platforms, raw answers, scores, and packaging — open only if you need the evidence behind a task.

Platforms & raw answers

Platform probes

Source evidence and raw answers for each assistant. Use this when a task needs proof — not as the default reading path.

Google Gemini
Gemini answers, which often inform Google’s AI surfaces and overview-style responses.
Demo adapter

Visibility

51

Accuracy

71

Sentiment

62

Share of voice

43

Query: What is Harbor Loaf? Explain what the company does, who it serves, and any notable facts.

Mentioned · primarySentiment neutralAccuracy accurateDemo response
Harbor Loaf (harborloaf.example) is a artisan bakery company. It sells products and services to teams that need a more dependable option than Tartine. Public write-ups are thin, so…

Peers named: Tartine

Source evidence

No URLs attached to this answer. Treat it as an unverified reconstruction until a cited public source exists.

Why the model might say this

  • This card used a labeled demo adapter, not a live model call. Treat it as a stand-in for how thin public records get reconstructed.
  • No citations were attached. Chat-style models often stitch training data, directories, and similarly named entities. A dated owned page (About, bio, /facts) is usually what changes later answers — not a prompt to “forget.”
  • Other names in the answer (Public, Typical, Independent) can bleed into this story when public sources are thin.
  • Website facts could not be fetched, so accuracy is only checked against the name and category you entered.
Raw answer
Harbor Loaf (harborloaf.example) is a artisan bakery company. It sells products and services to teams that need a more dependable option than Tartine. Public write-ups are thin, so descriptions sometimes collapse it into a generic artisan bakery name. Typical summaries mention customer support, implementation speed, and a narrower product focus than the category giants. Independent encyclopedic coverage is sparse, so most assistants reconstruct the story from the website, directories, and a handful of blog roundups.
Scores & gaps
68

Overall GEO score · Solid

Visibility
74
Weight 35%

Named in 20/25 answers (13 as the primary subject). Visibility is the mean of prominence: primary 100, featured 78, passing 42, absent 0.

Accuracy
77
Weight 25%

Accuracy maps flags to points (accurate 92, partial 62, unverifiable 55, not_mentioned 40, inaccurate 24). This run: 16 accurate, 5 not_mentioned, 4 partial.

Sentiment
64
Weight 20%

Sentiment is a keyword window around the brand mention (positive 88, mixed 58, neutral 70, negative 28, unknown 50). It is directional, not a substitute for human review.

Share of voice
51
Weight 20%

Share of voice ≈ brand-mentioning answers (20) ÷ (those + named competitor hits (19)). Category and alternatives queries weigh heavily here.

Formula: 0.35 × visibility + 0.25 × accuracy + 0.20 × sentiment + 0.20 × share of voice. Full methodology

Absent from “best of” category answers
critical
The name is missing when users ask who or what is notable in this field — a core GEO visibility gap.
ChatGPT · Claude · Gemini · Web AI
Trust answers are mixed or negative
high
Reputation-style prompts surface caution, complaints, or hedging instead of a clean trust narrative.
Web AI
Lost in head-to-head comparisons
medium
When users compare this name to a peer, some engines barely mention it — or default to the other party.
Gemini
Answers drift from claimed facts
high
Descriptions do not reliably match your site or category. Models may be confusing you with a similarly named entity or using stale training data.
ChatGPT · Claude · Perplexity · Web AI
History
Trend
Visibility, sentiment, and share of voice across completed runs for this brand.

Re-run or use Monitoring → Run scan now to plot a second point.

What changed
Compared with the previous completed run. Description flips fire when the same probe now says something different.

This is the first completed run. Watch and scan again to catch description flips, new peers, and accuracy drift.

Alerts

  • negative descriptionUnread

    Harbor Loaf: mixed or negative description on first run

    web-ai-search / Trust & reputation: …’t find a dense public trust record for Harbor Loaf. That usually means a small footprint, not necessarily a red flag — but it also means you should not take self-descriptions at

  • hallucinated descriptionUnread

    Harbor Loaf: accuracy issue on first run

    openai / Head-to-head comparison: Answer does not clearly reflect the claimed category “artisan bakery”.

Open monitoring
Goal & watch
Presence goal
CiteSight is about controlling AI/digital presence — not only getting cited more. Appear, correct, or fade from real public sources. We do not plant false claims about anyone else.

Presence goal

Win identity, category, and comparison answers with a consistent public record and citable pages.

Cadence & alert email
Start Monitor from the card above ($49/mo). This dialog only changes weekly/daily cadence or stores an email stub.

Optional. The Monitor card above is the product path.

We optimize real public presence. We don’t fake consensus or plant false claims about third parties. Correction work updates real public sources. Trust & limits

Packaging

Plans & packaging

Two commercial modes share one work queue. Switch the experiment flag per brand (or set PACKAGING_MODE) to A/B conversion later.

Default from env: modularBilling stub: none
Monitor subscription
Cadenced re-audits, trends, and in-app alerts. Improvement work is sold separately — pick actions below and fund them à la carte.

$49/mo

Fund these actions
Accept or skip each task in the work plan, then mark the ones you want CiteSight to execute. Checkout is per-action, not a bundle.

Cart is empty. On a task, click Fund execution to add it. Effort $29 / $79 / $149 (S/M/L).

Cart $0

No Stripe keys on this server — checkout records the plan locally, enables monitoring, and does not charge a card.