Everything inside a HiBot audit.
A HiBot audit is more than a score. It is a scored diagnosis, a prioritized action plan you can track, an evidence trail pulled straight from the AI engines, and a report built to circulate. Here is a breakdown of every component and related features.
The scorecard
Where you stand, in one number and six dimensions, tracked over time.
Score Dashboard
Your headline HiBot Score and the 6 ANSWER dimensions on one screen, each with a delta versus your last audit. Two rates sit alongside: how often the engines cite a source about you, and how often what they say is accurate.
- The 0 to 100 HiBot Score, with a verdict band from Invisible to Dominant.
- All 6 ANSWER dimensions, weakest flagged.
- Citation rate and accuracy rate at a glance.
Competitive Standing
We lead with the blunt outcome: how often AI names you, how often it names the category leader, and exactly where you rank among every brand you track. It used to be buried and Panorama-only. Now it headlines every report, on both plans.
- Your share, the leader's share, and your rank, N of M.
- The plain reading of it, in one sentence.
- On both Pulse and Panorama, whenever a competitor is tracked.
Coverage by Model
A brand can be strong in one engine and invisible in another. Coverage by Model shows the share of each engine's answers that actually name you, so you can see which models know you and which do not, out of the answers each one returned.
- Per-engine coverage, out of the answers that engine returned.
- Engines that returned nothing are excluded, not scored as zero.
- 3 engines on Pulse, 5 on Panorama.
Track it over time
Put a series on a recurring schedule and every report gains an over-time layer. AI answers are noisy, so these show movement lined up with dates, not proof of cause.
Dimension Trends
Each re-audit traces all 6 ANSWER dimensions across every run, not just the headline number, and drift flags call out the material moves and stalls so you do not have to hunt for them. Recurring audits also save 20% versus one-off.
- All 6 dimensions, traced across every cycle.
- Drift flags on the moves and stalls that matter.
- Since-baseline and vs-last deltas on each.
+17 points since April, driven by Appearance and Nomination gains.
Question-Level Changes
Question by question, see how you moved between runs: where you newly appeared, where you dropped out, where a competitor overtook you, and where your score rose or fell. It is the single most persuasive artifact an agency can put in front of a customer.
- Wins, losses, and competitor overtakes, per question.
- Prior run to this run, with the competitor named.
- A full compare view, down to each engine.
Accuracy Over Time
The claims log becomes a tracker across runs: every flagged claim is tagged new, persisted, corrected, or reappeared, with an accuracy-rate trend, so you can tell whether a known bad claim is still out there or finally gone.
- Claim states across runs, with corrected claims surfaced as wins.
- An accuracy-rate line across every cycle.
- Corrected means the engines stopped repeating it.
Timeline Notes
Log the changes you or your partners make, with dates, and they line up against your score history as pins on the trend. HiBot verifies nothing; any nearby movement in AI answers is a correlation, not a proven result.
- Dated notes pinned to the timeline.
- They persist across every run in the series.
- Correlation only, never a causal claim.
The action plan
The fixes, in priority order, with the evidence and the steps behind each one.
Impact-Effort Prioritization Matrix
Every recommendation, plotted by the impact it will have against the effort it takes. The quick wins sit top-left; the time sinks, bottom-right. It is the fastest way to decide what to do this week and what to leave for later.
- Four quadrants: quick wins, big bets, fill-ins, and time sinks.
- Each fix numbered, linked to its full detail below the fold.
- Recolors as you mark items complete.
ANSWER Deep Dive
A category-by-category read on each of the 6 ANSWER dimensions, written in plain language, with the exact customer questions we asked in that category. Every question links to the verbatim answers behind it, so nothing is a black box.
- A narrative for each dimension, strongest and weakest called out.
- The real questions, in your customer's language.
- One click through to the responses that produced the score.
In question after question about the category, the engines list competitors and move on without naming you. The same blind spot shows up across ChatGPT, Perplexity, and Google, so this is a category-recall problem, not one engine's quirk.
Tracked Strategic Recommendations
Every fix, ranked by priority and grounded in real quotes from the audit. Open a row to see the problem, the evidence the engines gave, the recommended steps, and the expected outcome. Set a status and keep notes right on the report, so it doubles as your working tracker.
- Priority, impact, effort, and horizon on every recommendation.
- Evidence quotes, tagged by the issue they show.
- Status and notes you can save, per recommendation.
| # | Recommendation | Imp | Eff | Pri | Status | |
|---|---|---|---|---|---|---|
| 1 | Break the unprompted-absence pattern Content · Marketing |
●●●●● | M | P1 | In progress | |
|
Evidence from your audit
“For KPI libraries I'd recommend Competitor B or Competitor C…” ChatGPT · omission Recommended steps
Expected outcome: lift Appearance from 17 toward 35+.
| ||||||
| 2 | Correct the enterprise-client claims Accuracy · Content |
●●●●● | S | P1 | Not started | |
| 3 | Seed accurate head-to-head content Content · SEO |
●●●●● | L | P2 | Complete | |
90-Day Roadmap
The recommendations, sequenced into fixed horizons with a score target for each. Every milestone and deadline drops into your calendar in one click, so the plan does not die in a PDF.
- 14, 60, and 90-day horizons, each with a score target.
- One-click Google Calendar and .ics reminders.
- Add a whole milestone to your calendar at once.
14 Days · Fix accuracy
60 Days · Earn recall
The evidence
The receipts behind every score: what the engines said, where they got it, and where you show up.
Audits in Your Customer's Language
If your customers ask AI in Slovak, an English audit measures the wrong market. Set your customer language once and every question is generated natively in it, phrased the way locals actually type, then run and captured in that language from first prompt to final answer. The report itself stays in English.
- Questions written natively in 26 languages, never translated word for word.
- The full audit runs in your customers' language: capture, scoring, and evidence.
- Language is locked into each run's methodology, like geography, so trends compare like for like.
Generated natively for each market, not run through a translator.
Where You Show Up
Instead of a linear list, a scannable grid: every question by every engine, each cell showing whether you were named, absent, or the engine returned no answer. Click any cell to read the actual response. On a re-audit, a compare mode overlays what flipped since last run.
- Named, absent, or no answer, at a glance.
- Click any cell through to the verbatim answer.
- A run-over-run compare shows what changed.
| GPT | PLX | GOO | CLA | GRK | |
|---|---|---|---|---|---|
| KPI library for startups | ✓ | − | ✓ | − | · |
| Templates by industry | − | − | ✓ | ✓ | − |
| Where analysts source metrics | ✓ | ✓ | ✓ | − | · |
Claims AI States as Fact Beta
Every claim an engine states about you that scoring flagged, checked against a snapshot of your own website and sorted by where it likely came from. A model error is something AI got wrong. A self-published claim is one AI repeated from your own site as fact, which is fine when your site proves it and worth fixing when it does not. A third-party figure came from another site. The fix is different for each, so the log tells you which is which. You get the final say on every entry.
- Each flag is verified against your own site and cited, or marked unverifiable, never presented with false confidence.
- Unproven claims from your own site are your fastest win: add the awarding body, the year, or a link, or soften the superlative.
- Re-sort any claim yourself. Only model errors count against your accuracy rate, and your HiBot Score never moves when you do.
- Nothing is deleted. Remove a false positive and restore it whenever you want.
Citations and Sources Log
Every source the engines cited when answering about you, parsed from the captured responses. It shows which domains the models trust in your category, so you know exactly where to earn a mention or correct the record.
- Every cited domain and URL, de-duplicated.
- Which engines leaned on each source.
- The question each citation appeared under.
Verbatim Responses Explorer
The full, word-for-word answer behind every score, captured by hand in a neutral session. Filter by category or engine, see the score and the reason for it, and read exactly what a real customer would have seen.
- Every response, captured verbatim, nothing paraphrased.
- The per-answer score and its rationale.
- Sourced, and filterable by category and engine.
For a startup, popular options include Competitor A and Competitor B, which offer ready-made templates… no mention of your brand.
Take it anywhere
We are deliverable focused. Export the report and collected responses in editable form, so you can easily tailor it for clients or your own purposes.
Word, PDF, and CSV Export
Take the report out of the browser. Download it as an editable Word document to add your own commentary, a print-ready PDF to circulate, or the raw results as a CSV to drop into your own analysis.
- Editable Word (.docx), with the matrix rendered as a diagram.
- A clean, print-ready PDF.
- Full results as CSV, every question and response.
.docx
.csv
White-Label Reports
Deliver AI visibility audits under your own brand. Add your report name and logo, and every HiBot reference is neutralized to a clean, white-label deliverable you hand to clients, exports included.
- Your logo and report name on every page and export.
- Neutral "AI Visibility Score" terminology, no HiBot branding.
- Print PDF and editable Word, both white-labeled.
MCP for AI Assistants
Your audit should not be trapped in a browser tab. Over MCP (the Model Context Protocol), the AI assistant you already use reads your delivered audits directly: scores, verbatim engine answers, the claims log, and trends. It can even record progress on each recommendation as your team works. Setup takes minutes on the developers page.
- Your agent queries your real audit data and logs what got fixed.
- API keys you mint, label, and revoke from your account.
- A free public server shares our methodology and research with any assistant, no account needed.
Read a finished report, start to finish.
We publish complete audits of real brands, Oura Ring, Costco, Accenture, Salesforce, and a redacted sample, in the exact format you receive. Download one before you buy.
See what AI says about you.
Every audit is run by hand by a human AI specialist, then scored and built into the full report and action plan above.