AI Visibility Tracker - Gemini Brand Mentions & Citations

AI Visibility Tracker — Gemini Brand Mentions & Citations

What does it do?

Weekly tracking: Reuse a monitor name in an Apify Schedule to compare AI answer mentions and mention order week by week. Monitoring uses the existing $0.12 per successful answer event; no extra event or price change.

Ask Gemini with Google Search grounding your buyers' questions and measure whether your brand appears, its mention order and cited URLs. Each usable answer produces one ANSWER row; successful queries cost $0.12 and failed queries are free. SUMMARY is saved to the key-value store and dataset. Perplexity support is prepared but requires an API key to be enabled by the operator; until then it returns a free engine_unavailable row.

How to use

1. Enter your brand, website and up to 10 competitors.
2. Add up to 10 buyer questions. Category questions without your brand name (for example, "Which CDN should a small business use?") measure whether an engine recommends you. If left empty, 10 questions containing your brand name are generated to inspect how the engine describes you.
3. Choose gemini (default) and optionally perplexity. Set samples to 1–5 to repeat each question on each engine.
4. Run on a paid Apify plan. Review brandMentioned, mentionPosition, citations and SUMMARY.shareOfVoice. For missed mentions, follow geoAuditUrl to the GEO readiness audit.

Example: {"brand":"Cloudflare","domain":"cloudflare.com","competitors":["Akamai","Fastly"],"queries":["Which CDN should a small business use?"],"engines":["gemini"],"samples":3}.

Weekly tracking

1. Enter a stable monitorName (for example, cdn-category-us) with your brand, domain, buyer questions and engines. The first run writes a baseline; its SUMMARY.changes is empty.
2. In Apify, save this input as a Task, then create a Schedule set to run that Task weekly. Keep the same monitor name and questions for comparable results.
3. Review SUMMARY.changes and CHANGES.md after each run. Optionally set webhookUrl to receive one JSON POST only when a change is detected. An unchanged run sends nothing.

Each question and engine has its own baseline in a named key-value store. Monitor names remain distinct even when their punctuation or capitalization would produce the same store slug; keep the exact same name across scheduled runs. A failed answer leaves that baseline intact. Added questions get a new baseline without a false alert. rank means mention order among your brand and listed competitors, not search-result rank. With multiple samples, a mention in any successful sample counts and the best observed mention order is compared. Share of voice is the brand's share of all monitored-name mentions for that question and engine; changes of at least 5 percentage points are reported. The regular ANSWER rows remain available for audit; the change digest only lists differences.

Input

• brand, domain (required): exact brand name and public website.
• competitors: up to 10 names.
• queries: up to 10 buyer questions; when empty, 10 brand questions are generated.
• engines: gemini (default); perplexity requires the operator's Worker API key.
• samples: 1–5 repeated answers per question and engine (default 1). Every successful sample is a separate engine query.
• monitorName: optional stable name for cross-run comparison. Empty means snapshot mode.
• webhookUrl: optional secret URL for one JSON change notification per run; used only with monitorName and nonempty changes.

Output

Each successful ANSWER has brand, domain, query, engine, sample, checkedAt, success, brandMentioned, mentionPosition, competitorMentions, citations, answerSummary, grounded, providerCostUsd, geoAuditUrl, and charged. Failed rows instead carry errorType and error. mentionPosition is the first appearance order among the monitored brand/domain and competitors in the answer text; it is not search-result rank. citations contains up to 20 source URLs. charged on a stored answer means charge eligible; confirmed charges are in the summary. providerCostUsd is the Perplexity API cost reported for that answer; Gemini has null.

SUMMARY has status, queriesAttempted, successfulQueries, failedQueries, brandMentions, visibilityScore, competitors, shareOfVoice, brand, domain, engines, samples, chargedEvents, notProcessed, estimatedGrossUsd, geoAuditUrl, unavailableEngines, availabilityNote, monitorName, and changes (plus firstRun, monitorStore, and webhook when applicable). visibilityScore is successful answers mentioning the brand divided by successful answers. shareOfVoice lists the brand and competitors with name, mentions and sharePercent; the denominator is all monitored-name mentions across successful answers, so one answer can count multiple brands. With no mentions, sharePercent is null. Failed samples are excluded. In monitor mode, changes contains query, engine, type, before, and after; types are new_mention, mention_lost, rank_rise, rank_drop, and share_of_voice_change. CHANGES.md contains the same digest. The webhook JSON contains monitorName, brand, domain, and changes.

Cost and limitations

$0.12 per successful engine query (engine-query), including each repeated sample. Failed, unavailable, skipped and free-plan queries are free. A spending limit stops further queries. The default maximum is 10 questions × 1 engine × 1 sample = $1.20; selecting both engines and 5 samples makes the theoretical maximum $12.00.

Live engine calls require a paid Apify plan. Free-plan runs return up to three paid_plan_required explanation rows without contacting either provider or charging. Perplexity stays unavailable until its key is approved and installed. AI answers vary; repeated samples help expose variation but do not represent a search ranking. Exact-name matching does not count nicknames or misspellings. The GEO audit predicts readiness and cannot guarantee citation.

FAQ

Am I charged when an engine fails? No. Only a saved, usable answer is charge eligible.

Can I monitor weekly? Yes. See Weekly tracking above.