For contracted clients we deliver a monthly brand visibility monitoring and implementation report, so the brand can spot changes at any time and adjust its optimization direction. Stop reporting with one-off screenshots — use trend data to explain what is happening.
L1 exposure → L2 citation → L3 conversion → L4 assets. Each tier builds on the last: miss one and you cannot explain where the results came from.
It answers the most basic question: does the brand appear in AI answers.
| Core KPI | Definition |
|---|---|
| Mention rate | Share of the core question set in which the brand is mentioned by AI |
| Top-recommendation rate | Share of answers in which the brand appears in the first recommended slot |
| Questions covered | Number of core questions in which the brand is successfully mentioned |
It answers the next question: is the AI saying the right thing, and are the sources authoritative enough.
| Core KPI | Definition |
|---|---|
| Citation accuracy | Share of statements that match the official fact base |
| Authoritative-source share | Share of cited sources that are authoritative media, industry associations or academic institutions |
| Positive-sentiment share | Share of AI statements that are positive or neutral |
| Hallucination resistance rate | Share of key parameters that AI does not misstate |
It answers what the boss cares about most: what business did this bring in.
| Core KPI | Definition |
|---|---|
| AI-channel search volume | Visits that arrive at the brand's own properties through AI engines |
| Qualified leads | Number of follow-up-ready opportunities coming from AI channels |
| Customer acquisition cost | Cost per qualified lead from AI channels |
| Brand-term index | Change in the search-and-discussion popularity index for brand terms |
It answers the most fundamental question: a year from now, what have we kept.
| Core KPI | Definition |
|---|---|
| Knowledge base size | Volume of knowledge assets that have been structured and tagged with metadata |
| Authoritative media coverage | Number of items picked up by authoritative media and vertical industry media |
| Content asset reuse rate | Share of a given asset that keeps being cited across multiple platforms and multiple models |
A fixed structure makes month-over-month comparison easy; conclusions come first so management can read it directly.
Which optimization actions and content output were completed this period: Schema rework, fact base updates, long-form articles and videos, source placement — all with timestamps and owners.
Month-over-month and year-over-year movement for every L1–L4 tier, with attribution: which actions moved which tier, and why anything fell short of target.
Competitors' mention share, top-slot ranking and content strategy shifts on the same question set, so you can see who is gaining and who is losing.
A next-cycle action plan built on the data: which question sets to add, which sources to strengthen, which pages to iterate.
Every sampling run uses a fixed question set, fixed phrasing and fixed model versions, so the data stays comparable.
Built on an industry core question set (usually 100–300 questions, sized to the company's scale and number of markets) and grouped into four categories — selection, comparison, scenarios and qualifications — so every sampling run is comparable.
Batch sampling on a single consistent basis across Doubao, DeepSeek, Kimi, ERNIE Bot, Qwen, Tencent Yuanbao, ChatGPT, Perplexity and others, reported per platform.
Automated sampling covers coverage and trends; human review judges accuracy (especially citation accuracy and hallucination resistance). Numbers are only released after the two cross-check.
Boundaries on what we promise, tiers on what we deliver, clarity on where the edges are — that is the precondition for dependable delivery.
We never promise "guaranteed number-one recommendation, 100% appearance, or a sales bet". Generative AI output is probabilistic, and promising a fixed ranking goes against how the technology objectively works.
No black-hat volume inflation, fake backlinks, fabricated media or unauthorized scraping. Every optimization action is white-hat and fully auditable.
We never fabricate data as client case studies. Every case figure we publish comes from a real project and is used with the client's authorization.
| Tier | Applies to | Typical scenarios | Response commitment |
|---|---|---|---|
| P0 Critical | Critical incidents | Website blocking · high-frequency major negative or incorrect citations · urgent regulatory notices | 30-minute response → 2-hour mitigation → 24-hour resolution |
| P1 Standard | Standard requests | Content optimization · platform errors · data anomalies · monthly report questions | Response within 1 business day, plan within 3 business days |
| P2 Advisory | Optimization suggestions | Strategy iteration · new platform adaptation · knowledge graph expansion · team training | Planned fortnightly or monthly, synchronized in the regular review meeting |
Four tiers of metrics. L1 exposure (mention rate / top-recommendation rate / questions covered, ≥45%, weekly). L2 citation (citation accuracy / authoritative-source share / positive-sentiment share / hallucination resistance rate, ≥98%, weekly). L3 conversion (AI-channel search volume / qualified leads / customer acquisition cost / brand-term index, ROI ≥3:1, monthly). L4 assets (knowledge base size / authoritative media coverage / content asset reuse rate, compounding, quarterly).
Four fixed sections: work delivered this month, metric movement analysis, competitor monitoring and next month's optimization strategy. The goal is to let clients spot changes at any time and adjust their optimization direction — not to hand over a single screenshot.
P0 critical incidents: 30-minute response, 2-hour mitigation, 24-hour resolution. P1 standard requests: response within 1 business day, plan within 3 business days. P2 optimization suggestions: planned fortnightly or monthly, synchronized in the regular review meeting.
Three. We never promise "guaranteed number-one recommendation, 100% appearance, or a sales bet". We never use black-hat volume inflation, fake backlinks, fabricated media or unauthorized scraping. We never fabricate data as client case studies.
Different metrics move and respond at different speeds. L1 and L2 shift quickly after a model index update, so they need weekly sampling. L3 follows the sales cycle, so monthly is the right cadence. L4 is a long-term accumulation, so quarterly observation is more meaningful. Reporting at separate cadences stops short-term noise from driving long-term judgments.
aileo provides anonymized diagnostic samples and monthly monitoring report samples, available on request during a sales conversation.