AI Brand Visibility Auditor — Profound AI Alternative

Audit how selected model APIs mention your brand in supplied prompts. Export answer excerpts, brand and competitor counts, parsed links, model identifiers, and timestamps per brand-prompt-model observation. For AI-visibility analysts, each dataset row represents one brand, prompt, and model-API answer observation.

Workflow: put the results to work

Choose the brand, competitor names, and exact prompts you want to inspect. Compare the returned answer-text observations while keeping model and prompt context visible. These snapshots describe API responses, not real-user prompt demand or every consumer chatbot experience.

What this Actor does

The Actor sends supplied prompts to selected managed model APIs: openai, perplexity, or gemini. It then applies deterministic exact-name matching to a target brand and any competitor names you supply. URL fields are parsed only from the returned answer text and are clearly labeled as such.

The returned model API answer can differ from a consumer answer-engine product, including its product-specific search, personalization, region, subscription tier, tools, and response-time behavior. Treat every row as a source-labeled API snapshot at its collectedAt time—not as a universal visibility score or a claim about what every user sees.

When to use it

• You have a focused prompt set and need a current, structured answer-text snapshot.
• You want to schedule a repeatable brand-versus-named-competitor check.
• You need API-ready JSON with prompt, model, collection-time, and explicit match semantics.
• You need a bounded workflow that can feed your own reporting, alerting, or review process.

When not to use it

• You need real-user prompt-volume data, historical answer-engine reporting, or a market-wide panel dataset.
• You need a broad platform with dashboards, collaboration, content agents, or crawler and traffic analytics.
• You need independently verified citations or a universal claim about consumer answer-engine behavior.

Scope and complementary workflows

• Profound fits licensed real-user prompt volumes, demographic modeling, or market-wide prompt demand.
• Profound also fits multi-month historical databases, dashboard collaboration, content-generation agents, and server/CDN AI crawler analytics.
• Pair this Actor with human review when prompt discovery, competitor discovery, external URL verification, or citation assessment is required.
• Use the selected model API route when an identified API response is the desired measurement surface.

Start with this Actor for answer-text visibility, then use AI SEO Tracker for a site-level technical audit of crawlability, metadata, robots, sitemap, and llms.txt. AI SEO Tracker audits public web properties while this Actor focuses on model API answers.

Input

brandName and prompts are required. Keep the questions specific to the customer decision you want to inspect. Add brandDomain only when you want to flag parsed answer URLs that match that domain; it is not an independent citation verifier.

{
  "brandName": "Example Analytics",
  "brandDomain": "example.com",
  "competitors": ["Rival Analytics", "Another Analytics"],
  "prompts": [
    "Which analytics platform is best for a small B2B SaaS team?",
    "What are practical alternatives to Example Analytics for product analytics?"
  ],
  "engines": ["perplexity", "gemini"],
  "maxResponses": 4,
  "responseFormat": "concise"
}

maxResponses caps all prompt × engine combinations before model requests begin. Use dryRun: true to validate a scheduling or API payload without requesting model answers or charging the snapshot event.

Output

Every dataset record is a single model API response, not an aggregated dashboard report.

Field	Meaning
prompt, engine, model	Exact supplied prompt and selected provider model route.
responseSource	Always model_api; this is the data-provenance boundary.
brandMentioned, brandMentionCount	Exact-name match result in returned answer text.
brandShareOfVoice	Brand exact mentions ÷ (brand + supplied competitor exact mentions), or null when neither appears.
competitorsMentioned, competitorMentionCounts	Names from your list and their exact match counts.
sourceUrls, citedBrandUrls	URLs parsed from answer text; brand URLs are only hostname matches to brandDomain.
mentionContext, sentiment	Bounded text near the first mention and a simple keyword heuristic—not a human sentiment verdict.
answerText, collectedAt	Returned answer text and ISO 8601 collection timestamp.

Example abbreviated record:

{
  "recordType": "ai-answer-visibility-snapshot",
  "brandName": "Example Analytics",
  "prompt": "Which analytics platform is best for a small B2B SaaS team?",
  "engine": "perplexity",
  "model": "perplexity/sonar",
  "responseSource": "model_api",
  "brandMentioned": true,
  "brandMentionCount": 1,
  "brandShareOfVoice": 0.5,
  "competitorsMentioned": ["Rival Analytics"],
  "sourceUrls": ["https://example.com/docs"],
  "citedBrandUrls": ["https://example.com/docs"],
  "collectedAt": "2026-08-31T12:00:00.000Z"
}

OUTPUT always contains the terminal outcome, counters, billing counts, and warnings. RUN_SUMMARY adds parsed input, individual model-attempt diagnostics, start/finish times, stop reason, and cost estimate. Valid outcomes are COMPLETE, PARTIAL, VALID_EMPTY, INVALID_INPUT, UPSTREAM_FAILED, and CONFIG_ERROR.

API example

Replace the placeholder token after the Actor is privately deployed. Keep Apify tokens in an authorization header or secure environment variable rather than a URL or shared dataset.

curl -X POST "https://api.apify.com/v2/acts/khadinakbar~profound-ai-alternative/runs" \
  -H "Authorization: Bearer $APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "brandName": "Example Analytics",
    "prompts": ["Which analytics platform is best for a small B2B SaaS team?"],
    "engines": ["perplexity"],
    "maxResponses": 1
  }'

Use the returned run's default dataset ID to retrieve the structured rows. The hosted MCP contract has been exercised by a Codex client: it discovered the typed Actor tool, completed a bounded run, and retrieved the persisted structured row. That path does not imply a universal client compatibility or model-output guarantee.

Agent prompt card

Run a bounded current AI-answer snapshot for [brand]. Use these exact prompts: [prompts]. Compare only these named competitors: [competitors]. Query [engines], cap the run at [maxResponses] snapshots, and return the dataset rows plus OUTPUT. Treat the result as a source-labeled model API snapshot, not real-user prompt-volume data or a universal visibility score.

How this workflow compares with Profound

Profound is the better fit when you need its broader platform: real-user prompt-volume data, historical reporting, multiple answer-engine coverage, dashboard collaboration, agents, content workflows, or AI crawler/traffic analytics. Its official pricing page currently lists a recurring Starter plan with a different included scope.

This Actor is only an on-demand alternative for the narrow workflow of sending your supplied prompts through selected model APIs and exporting a timestamped structured snapshot. It does not make a lower-cost, faster, more reliable, or full-suite-replacement claim against Profound. The private ./ALTERNATIVE_ACTOR_DOSSIER.json records the dated scope and pricing evidence; a same-job cloud-run cost comparison remains unverified.

Decision area	This Actor	Profound / current boundary
Workflow scope	One supplied brand, prompts, selected model APIs, and structured response rows.	Broader answer-engine visibility platform and reporting workflow.
Prompt volumes	Does not provide real-user prompt-volume data.	Its official materials describe licensed opt-in panel prompt data.
Output provenance	Each declared row stores prompt, model, responseSource, and collectedAt.	Its documented reporting API returns platform-owned visibility report data.
Input friction	Requires a brand name and one to ten supplied prompts; it does not use a Profound account.	Its API works against its own authenticated data model.
Cost	Uses an event price per persisted snapshot, plus separate platform usage; final validation reconciles expected snapshots with primary events, but no all-in same-job comparison is claimed.	Uses a different recurring-plan scope; no same-job allocation has been measured.
Effective efficiency	A bounded snapshot workflow; no same-workload throughput or active-review-time comparison is claimed.	No authorized same-job comparison was run.
Automation	The same saved input can be rerun; a live scheduled-workflow evaluation is pending.	Broader platform automation is outside this Actor's tested scope.
Reliability	Final validation covers normal, diagnostic, cap, and selected-model paths; it is not a production reliability guarantee.	No same-job reliability test was performed.

Best results and workflow guidance

• Keep brand and competitor names unambiguous. Exact matching will not infer aliases, subsidiaries, or pronouns.
• Use realistic customer questions. The Actor does not claim they represent real consumer prompt volumes.
• Repeat the same saved input on a schedule if you want a buyer-owned time series; compare rows only when engine, prompt, response format, and named competitor list are stable.
• Review answerText and mentionContext before acting on the calculated fields. They are mechanical observations, not research conclusions.
• Treat parsed URLs as answer-text observations and confirm source attribution in the relevant answer-engine or source workflow when that decision matters.

Responsible use and data handling

The Actor sends supplied prompts to managed model APIs and stores returned answer text, calculated fields, and terminal diagnostics in your Apify storage. Submit only data you are authorized to use, and follow the applicable Apify and model-provider terms. This is a read-only monitoring workflow; it leaves brands, websites, models, and third-party accounts unchanged.

Builder's note

I designed this Actor to favor a narrow, reproducible record over a vague overall score. My goal is for every row to preserve the question, selected model API, returned text, exact matching rule, parsed URLs, and collection time so a human or downstream system can inspect the observation in context.

Current evidence status

The private release candidate has local tests, static schema/readiness checks, live pricing and metadata readback, a final cloud acceptance matrix with output and charge reconciliation, and a hosted-MCP client run with structured dataset retrieval. The private release-evidence log is the authoritative audit record; the Store page makes no production uptime, accuracy, cost-comparison, or full-suite-replacement claim.

Pricing and run costs

This Actor uses Pay per event plus Apify platform usage. The Pricing tab lists the current event rates and billing terms.

Event	Billing unit	When it applies
apify-actor-start	Actor Start	Charged when the Actor starts running. Number of events charged depends on Actor memory (one event per GB, minimum one event).
ai-answer-snapshot	AI answer visibility snapshot	Charged once for each validated model API answer snapshot persisted to the dataset.

Run cost combines the charged events and Apify platform usage. Review the run charge limit and requested result count before starting.

Independent alternative

This Actor provides the specific workflow described above. It is not affiliated with or endorsed by Profound AI; the named product and its trademarks belong to their respective owners.

Connect an AI agent

Use the Apify MCP configurator to choose an available client connection. Inspect this Actor’s current input schema and required credentials before running it.