# Reputation

> Source: https://www.aisyndicate.com/reputation/
> Provider: AI Syndicate · Last updated: 2026-07-17
> Markdown version for LLMs and AI agents. Canonical HTML: https://www.aisyndicate.com/reputation/

## What is Reputation?

**Reputation** is AI Syndicate's reputation monitor for the AI layer. Buyers no longer read twenty reviews and average them in their heads — they ask AI "what do people think of this company?" and get one synthesized verdict. Reputation asks **13 AI surfaces** that exact question, scores the tone of every actual answer — **positive, neutral, negative** — from both **training memory and live retrieval**, extracts the recurring themes in the engines' own words, and tracks ratings across the review platforms those answers draw on — each proven the brand's own before it counts.

Brand Intel asks whether AI knows who you are. Reputation asks what it thinks of you.

## The 13 surfaces

Ten chat and voice engines (ChatGPT, Claude, Gemini, Perplexity, DeepSeek, Meta AI, Grok, Mistral, Siri, Alexa+) plus Google AI Overviews, Copilot, and Reddit. For the non-chat surfaces, the verbatim text each one shows about the brand is fetched and its tone is scored from that text only.

## What it measures

1. **AI sentiment across 13 surfaces** — each answer's tone scored positive/neutral/negative, averaged only across engines that actually recognise the brand.
2. **Training vs live retrieval** — every chat engine answers twice: from training memory, then grounded with live web context (the way retrieval-backed products actually answer). Both tones are shown per engine, plus the gap between them.
3. **Themes, in their words** — up to five recurring themes extracted from the engines' own summaries, each attributed to the engines saying it, judged only from scan text.
4. **AI memory map + get-known playbook** — which engines know the brand from training, which only find it with web context, and which can't place it at all, with a per-engine playbook of fixes.
5. **Review platforms, tracked** — Yelp and Google ratings with real review excerpts, plus G2, Capterra, TrustRadius, and Product Hunt ratings — each proven the brand's own before it counts, with 30-day deltas per platform. Reddit threads included as what AI reads.
6. **Rival benchmark + alerts** — tone benchmarked against up to five competitors, a Monday auto-refresh, and a tone-shift email alert.

## Honest by design

- **Unknown ≠ neutral.** An engine that can't place the brand is excluded from the average and listed by name — never counted as a neutral vote. It gets a second, web-grounded pass; if retrieval finds the brand, that tone is shown with a web-context tag.
- **Failures are failures.** A surface that errors is retried, then reported as a scan problem — never counted as an opinion.
- **Provably yours, or not counted.** Ratings come from licensed and public data — never scraping — and a profile only counts once it's proven the brand's own; a same-name stranger's rating never leaks in. The exact sources and matching pipeline stay ours.
- **Every tone is attributed.** Per-engine bars, per-mode (training vs retrieval), themes quoted from the engines' own summaries.

## How it works

1. **Ask** — 13 surfaces get the question buyers ask; each chat engine answers twice (training memory, then live web retrieval).
2. **Read** — the tone of each actual answer is scored positive/neutral/negative and normalized. Unknowns are excluded and named; failures are retried, then reported as failures.
3. **Compare** — per-engine and per-mode tone bars, themes in the engines' own words, six review platforms with 30-day deltas, and a tone benchmark against up to five rivals.
4. **Act** — negative themes become counter-briefs in Content with one click; unknown engines get a get-known playbook; a Monday auto-refresh and tone-shift email alerts keep watch between scans.

## Reputation vs. Brand Intel vs. Hallucination Watch

- **Brand Intel**: does AI know WHO you are — identity and confusion?
- **Hallucination Watch**: are AI's FACTS about you right — with corrections drafted?
- **Reputation**: what does AI THINK of you — the tone, the themes, and the review platforms feeding them?

## FAQ

**What is Reputation?** AI Syndicate's AI-layer reputation monitor: 13 surfaces asked what people think of your brand, every answer's tone scored (training + live retrieval), themes extracted in the engines' own words, ratings tracked across six review platforms plus Reddit, a rival benchmark, and tone-shift alerts.

**How is AI sentiment measured?** From the actual answers: each surface is asked "What do people think of your brand?" and the tone of that answer is scored positive/neutral/negative. The aggregate averages only engines that recognise the brand; the exact scoring recipe stays ours.

**What happens when an AI engine doesn't know my brand?** Honest exclusion: it's named, never counted as neutral, given a web-grounded second pass, and mapped with a per-engine get-known playbook.

**What's the difference between training tone and retrieval tone?** Training tone comes from the model's memory alone (which can lag by years); retrieval tone is the same engine grounded with live web context — how shipping products answer. Both are shown per engine.

**Where do review ratings come from?** Licensed and public rating data — never scraping — and a profile only counts once it's proven yours. Reddit threads appear as mentions, never as a synthesized rating. The exact sources and matching pipeline are part of our moat.

**Which plans include it?** Every plan — on-demand scans with a 24-hour cache, Monday auto-refresh, and tone-shift email alerts included.

**Can I benchmark against competitors?** Yes, up to five rivals on the same question — with an honest "AI can't place them" when an engine doesn't know a rival.

## Glossary

- **AI sentiment** — the positive/neutral/negative tone split of an AI surface's actual answer — measured from the answer itself.
- **Training tone** — an engine's tone from trained memory alone, no browsing.
- **Retrieval tone** — the same engine's tone grounded with live web context.
- **Unrecognized** — an engine that can't place the brand; excluded from the aggregate and listed by name.
- **Reputation theme** — a recurring claim extracted from the engines' own summaries, attributed.
- **Tone-shift alert** — the email sent when aggregate tone moves meaningfully (negative +8 or positive −10 points).

## Related

- [Brand Intel](https://www.aisyndicate.com/brand-intel/)
- [Hallucination Watch](https://www.aisyndicate.com/hallucination-watch/)
- [AI Access](https://www.aisyndicate.com/ai-access/)
- [Security Access](https://www.aisyndicate.com/security-access/)
- [All products](https://www.aisyndicate.com/products/)