MEIKAILoading
Loading
Brand Visibility
Large Language Models like ChatGPT and Gemini are becoming the first layer of brand discovery. Meikai tracks how your brand is mentioned, described, and recommended, and how that compares to your competitors.
37%
of search now starts with AI
Shortlist
of brands named in each AI answer
14+
LLMs monitored continuously
Understand how often your brand appears in AI responses across key topics and prompts. If your brand isn't named in the answer, buyers who rely on it may never see you.
Meikai runs structured prompt sets across leading LLMs, captures brand mentions, rankings, and citations, and benchmarks performance over time.
AI doesn't just mention brands; it frames them. Narrative positioning influences trust, authority, and purchase intent.
We compare your brand against key competitors across attributes and product categories, analysing how leading LLMs describe, rank, and recommend brands.
NikeAn AI answer names only a short list of brands. If you're not on that shortlist, you're excluded from the most influential moment in the customer journey.
We analyse competitive visibility trends, presence distribution, share of voice by topic, and competitor-exclusive citations.
Nike
Adidas
New Balance
ASICS
Under ArmourA useful first audit answers three questions: whether your brand appears at all in the answers your buyers get, how often it appears relative to named competitors, and whether what is said about it is accurate. Manual spot checks cannot answer any of those reliably, because a single run of a prompt tells you nothing about how the model behaves across repeated runs.
The audit runs a standardised prompt set across the models your market actually uses, repeats it, and reports presence, position and share of voice against the competitors you name. Comparing two or three rivals on an identical prompt set is more informative than a score of your own, because it shows which competitor holds the position you want and which sources put them there.
Accuracy matters as much as presence. A model that recommends you while misdescribing a product does more damage than one that omits you, so the audit records how your brand is characterised, not only how often it appears.
Run it again a week later. A vendor who cannot explain the movement is itself the finding.
The spread column is the reason for repeated runs. A single Gemini query could have returned any answer within 11 points of this one.
Leadership does not need prompt logs. Four figures carry most of the meaning: share of voice against competitors, how often you appear in the top few recommendations and not merely mentioned, how your brand is framed, and which domains the models cite when they describe you.
Report those on the same prompt set each month so the comparison is real. Month-on-month movement on a stable set is a signal you can act on, whereas movement on a prompt set that changed between reports tells you very little. Where a change was deliberate, show the before and after, and be explicit about which part is measured and which part is inference.
Built for a standing monthly slot, not a deck assembled once and never repeated.
Share of voice
vs three named competitors
Leading recommendations
named in the first three, not just mentioned
Framing
answers describing the brand favourably
Citation concentration
share of answers resting on one domain
Four figures, one page, no prompt logs. Movement only means something because the prompt set was the same in both months.
Meikai monitors 14+ AI engines: ChatGPT, Gemini, Google AI Mode, Google AI Overviews, Perplexity, Claude, Grok, Copilot, Wenxin, Qwen, Doubao, DeepSeek, Naver AI Tab and Alexa. The list grows as new assistants reach real usage. How many of them a given customer tracks depends on the plan: Starter covers ChatGPT only, Growth and Enterprise each include six engines chosen from the full set, and a custom plan can cover all of them.
GEO, with the SEO work that AI answers depend on. The two are measured differently. Google AI Mode and AI Overviews draw on the Google Search index, so classic crawlability, structure and SEO work carry over to them directly. The other engines answer from their own retrieval and training, where visibility is earned through source influence rather than search ranking. Site Scanner audits how AI crawlers reach your site, covering the access policy for GPTBot, OAI-SearchBot, ClaudeBot, Google-Extended, PerplexityBot and Googlebot. That crawler list is deliberately shorter than the engine list, because not every assistant we monitor publishes a documented crawler. Crawler activity is not limited to it: the CDN logs report every bot that reaches your site, with IP verification for OpenAI, Perplexity, Google and Bing. Meikai does not report Google organic rankings. Search Console is a separate connection: you sign in with Google, pick the property, and the search performance data is ingested on a schedule alongside the crawler and referral views. It is rolling out and switched on per brand, so ask us where your account stands.
Meikai runs structured prompt sets across major LLMs and measures brand mentions, ranking position, recommendation frequency, and citation sources.
It measures how often your brand appears in AI-generated answers compared to competitors across relevant prompts.
Yes. Brand perception scores the attributes engines associate with you, shown as a radar and a distribution, tracked over time and compared against competitors by topic and category. Every rating opens to the answer that produced it, the reason the engine gave, and the sources it cited.
Yes. Meikai analyses sentiment and narrative framing inside LLM responses, so you understand not just whether you appear, but how you are described.
Yes. Each cited page is classified for how it treats your brand: whether the brand is the main subject, a substantial part of the page or a passing mention, whether the portrayal is positive, neutral, negative or mixed, what kind of page it is, from ranked list and comparison to review, guide, product page or forum thread, and whether it is editorial, sponsored, affiliate or owned content. You can correct any of those classifications, and the corrected value is what reporting and exports use.
Yes. Brand Accuracy measures the claims AI answers make about your brand against a governed truth set: the facts and canonical sources you approve, versioned, with the active version read-only. Each claim is assessed against the facts that were valid on the day of the answer. The result is a hallucination rate and a factual-inaccuracy rate per platform, with confidence ranges, held as directional until enough claims have been assessed. Contradictions land in an incident queue with the evidence frozen, where your team confirms, rejects or defers each one.
Often, yes. A single answer typically names a shortlist rather than a full page of results, so being on it is the new competitive frontier.
A standardised prompt set run repeatedly across the models your market uses, reporting presence, position, share of voice against named competitors, and how accurately your brand is described.
By running an identical prompt set for you and the competitors you name, then comparing recommendation frequency, position and the sources cited for each. The comparison only means something if everyone is given the same prompts.
Share of voice against competitors, how often you appear in the leading recommendations and not only mentioned, how your brand is framed, and which domains are cited. Report them on a stable prompt set so month-on-month movement means something.
Explore our next solution
Make your website AI-readable with the Site Scanner.