Multi-Language Analysis
Track brand visibility across language markets with localized recommendations and content guidelines.
Overview
Multi-Language Analysis detects which languages appear in your AI snapshots and reveals how your brand visibility varies across language markets. It provides priority-ranked recommendations and localized content guidelines for each detected language.
As AI models serve global audiences, your brand may appear differently in English, Spanish, Japanese, and other languages. This feature helps you identify gaps and opportunities.
How Language Detection Works
The system uses a two-tier detection approach:
- Non-Latin scripts — Unicode character ranges detect CJK (Japanese, Korean, Chinese), Arabic, Hindi (Devanagari), and Russian (Cyrillic) with high accuracy.
- Latin-script languages — Common word-frequency patterns distinguish Spanish, French, German, Portuguese, Italian, Dutch, Turkish, Polish, and Swedish from English. English is scored on its own high-frequency words too — and a Latin-script language is only assigned when its word-frequency score clears a minimum and beats English by a clear margin. This prevents long English passages that happen to contain a few words shared with Romance languages (e.g. "per", "do", "la") from being mislabeled.
Supported Languages (16)
| Language | Code | Detection Method |
|---|---|---|
| English | en | Word frequency |
| Spanish | es | Word frequency |
| French | fr | Word frequency |
| German | de | Word frequency |
| Portuguese | pt | Word frequency |
| Italian | it | Word frequency |
| Dutch | nl | Word frequency |
| Japanese | ja | Unicode (Hiragana/Katakana/Kanji) |
| Korean | ko | Unicode (Hangul) |
| Chinese | zh | Unicode (CJK Unified) |
| Arabic | ar | Unicode (Arabic block) |
| Hindi | hi | Unicode (Devanagari) |
| Russian | ru | Unicode (Cyrillic) |
| Turkish | tr | Word frequency |
| Polish | pl | Word frequency |
| Swedish | sv | Word frequency |
What You Get
The page opens with one sentence — how many languages the answers came in, and what share of runs the primary language is — and then one sortable table.
Language Breakdown
One row per detected language:
| Column | What it shows |
|---|---|
| Runs | How many captured answers (runs) were written in that language. |
| Named | How many of those runs named your brand, as "n of N" with the share, drawn as a bar whose width is the share and whose count sits on the bar — so a language with 4 runs never draws like one with 400. Beside it, the credible interval on the chip's own rules: "low confidence" under 10 runs, "below target sample" under 30, otherwise the half-width in points ("±x pts"). |
| Tone | The leading verdict — Positive, Neutral, Negative, or Mixed when two tie — over the runs that named your brand and were judged, with that count beside it ("Positive · 12 judged"). Under four judged answers only the count is shown; nothing judged reads "not judged". Never a −1…1 average. |
| Competitors named | How many of the language's runs named any tracked competitor, as "n of N". |
Every column sorts. Bars are one hue; a low share is not coloured red, because a low share on a market you have not localised for is expected, not an alarm.
If only one language is detected there is no cross-market comparison to make; the table still shows the one row and the page says how to add a market (Settings → Locations).
Retired on 9 September 2026
Three headline figures were removed from this page and their definitions are marked retired in the Metric definitions appendix:
- Localization Readiness (0–100) — blended the mention rate ×1.5, a rescaled sentiment mean and
min(100, runs × 10), so capturing more runs raised "readiness" with no change in the market. The per-language table above is what it summarised. - Best Mention Rate — the maximum across languages, a cherry-picked headline; the gap is what the page exists to show, and the table shows it.
- High Priority — only the number of recommendation cards carrying that badge, restated above the cards.
Language-Specific Recommendations
Priority-ranked recommendations for each language, sorted by urgency:
- High priority — Languages where your brand has low mention rates or is losing to competitors.
- Medium priority — Languages with moderate presence but room for improvement.
- Low priority — Languages where visibility is already strong.
Each recommendation includes:
- Action items — Specific steps to improve visibility.
- Content guidelines — FAQ structure, schema markup notes, and content tips tailored to that language.
Getting Started
- Create multilingual prompts — Add prompts in your target languages, or configure Location settings with different language codes.
- Capture snapshots — The system automatically detects languages from response content.
- Open Language coverage — Navigate to the page from the sidebar (Performance › Language coverage). The same runs read by question type instead of by language are on the Performance review.
- Follow recommendations — Start with high-priority languages.
Content Guidelines by Language
The system provides language-specific advice for the 6 most common content markets:
- English — Focus on clarity, conciseness, and FAQ schema.
- Spanish — Consider regional variants (es-MX vs es-AR); FAQ with "Preguntas frecuentes".
- French — Distinguish fr-FR and fr-CA; use hreflang tags; formal tone default.
- German — Handle compound words; "Sie" formality; capitalize nouns in schema.
- Japanese — Consistent kanji/hiragana usage; FAQ as "よくある質問"; katakana for foreign terms.
- Chinese — Simplified vs Traditional distinction; pinyin/phonetic guides; platform-specific considerations.
Plan Requirements
Multi-Language Analysis requires the Advanced AI Insights feature, available on Pro-Individual plans and above.