About Beacon
Customer decisions now run through AI. Beacon runs a brand's customer journeys through AI assistants such as ChatGPT, Gemini, Claude, Perplexity and Google AI Mode, works out why AI chooses a rival, and ranks the changes that win the decision back. Beacon works on the public AI assistants and on inputs you provide. It does not connect to your customer database, your CRM, your analytics, or your production systems.
Beacon is always the questioner. We do not observe your customers' conversations with AI assistants.
How Beacon works
Questions
Beacon poses its own questions to public AI assistants, including ChatGPT, Gemini, Claude, Perplexity, Grok and Google AI Mode, written the way your buyers ask them.
Analysis
The answers flow into language-model analysis: claim extraction, source attribution, and scoring of awareness, comparison and endorsement. These are stateless inference calls. Customer data is not used to train any model at any stage.
Outputs
- The dashboard. A read-only view of how AI describes your brand across each assistant.
- Research by Bee. Bee, Beacon's research agent, investigates specific findings on demand; results are added to a brand-specific library reviewable by your team.
- The playbook. Recommendations your content and web teams implement on your owned properties, ranked by impact. Beacon authors copy and structural guidance; your teams ship it.
Data provided
| Input | Detail |
|---|---|
| Brand identity | Name, primary and supplementary domains, voice and positioning |
| Strategic direction | Products to win on, buyer personas, priority competitors |
| Optional context | Known content gaps, target opportunities, prior research |
Data accessed
| Source | Detail |
|---|---|
| Public AI assistants | Answers to the category and brand questions Beacon poses |
| Public web content | Brand and competitor URLs fetched as a search engine would |
| Dashboard accounts | Names, emails, role assignments. Auth via Clerk; passwords not stored |
Data not accessed
- No consumer PII. Names, emails, phone numbers, addresses, account identifiers, or any other personal data of your customers.
- No payment data. Card numbers, banking details, transaction records. Beacon is not in PCI scope.
- No consumer IP addresses, cookies, session identifiers, or device fingerprints.
- No connection to your systems. No read or write access to your CRM, marketing automation, analytics, ad accounts, CMS, ticketing, or production databases.
- No write access, anywhere. Playbook recommendations are changes your teams implement; Beacon never pushes, posts, or modifies anything in your environment.
- No PHI. Beacon is not configured to receive Protected Health Information.
Models in use
Beacon uses third-party language models for analysis. We use the latest stable model from each provider. No model provider Beacon uses trains models on Beacon's API traffic, and Beacon does not fine-tune or train models on customer data at any stage.
| Provider | Model families |
|---|---|
| Anthropic | Claude Opus 4, Claude Sonnet 4, Claude Haiku 4 |
| OpenAI | GPT-5.4, GPT-5.4 Pro, GPT-5.3 |
| Gemini 3, Gemini 2.5 | |
| xAI | Grok |
| Perplexity | Sonar |
| Xiaomi | MiMo (analysis only, via OpenRouter) |
| DeepSeek | DeepSeek V3 (analysis only, via OpenRouter) |
| MiniMax | MiniMax M2 (analysis only, via OpenRouter) |
Sub-processors
Beacon uses the sub-processors below to deliver the service. Each operates under its own published data-processing and security terms. We update this list as our sub-processors change.
| Sub-processor | Purpose | Region |
|---|---|---|
| Amazon Web Services (RDS, S3, ECS) | Primary database, object storage, and container compute (background workers) | US (us-east-1) |
| Vercel | Application and website hosting, and object storage (Blob) | US |
| Anthropic | Claude API for analysis, and the free audit form's focus suggestions | US |
| OpenAI | GPT API for analysis | US |
| Gemini API for analysis | US | |
| xAI | Grok API for analysis | US |
| Perplexity | Sonar API for analysis | US |
| OpenRouter | Model-routing gateway for cost-optimized analysis models; restricted to providers that do not train on or retain request data | US |
| Inngest | Background workflow orchestration | US |
| Clerk | Authentication for the dashboard | US |
| Resend | Transactional email to dashboard users, and website form messages to Beacon's team | US |
| PostHog | Product analytics for the website and the dashboard (dashboard user email, page views, session replay) | US |
| Slack | Internal operational alerts and usage summaries for Beacon staff | US |
Beacon also uses providers of hosted browsers and search APIs to put its questions to AI assistants. We share their names with customers under a non-disclosure agreement.
Optional integrations
The integrations below are included in the Beacon license at no additional fee. They are independent of the baseline product, and the baseline product does not depend on any of them.
Each optional integration is disabled by default. Enabling any integration requires written authorization signed by an authorized customer representative. Customers can specify in the master agreement which integrations are permitted; integrations outside that list are blocked at the contract and ingest level. A business-side request cannot unilaterally introduce an unreviewed data path.
Each enabled integration can be technically disabled at any time. Beacon's ingest layer rejects data from any integration not currently authorized.
Anonymized CDN log integration
Correlate AI assistant referral traffic with your owned properties via a narrow slice of CDN logs.
- Fields received
- Timestamp, URL path and host, User-Agent
- Fields not received
- IP addresses, cookies, session IDs, auth headers, request or response bodies, geo, device fingerprints
On Cloudflare Logpush, Akamai DataStream 2, and Fastly Real-Time Logs, field selection is opt-in at the edge: fields not explicitly included in the stream are never transmitted to Beacon. As a defense-in-depth control, Beacon's ingest layer discards consumer IP addresses, cookies, session identifiers, and authorization headers before persistence, regardless of whether the source CDN transmits them.
Supported delivery: Cloudflare Logpush, Akamai DataStream 2 (mTLS), or Fastly Real-Time Logs.
Beacon JavaScript tag for agent traffic
A site-installed tag that classifies the User-Agent against known AI bot patterns and pushes only matching events to Beacon. One-way push from your site; no read path back. Removable at any time; scope restrictable at the contract level.
- Fields pushed
- Timestamp, URL path, User-Agent, referrer
- Fields not pushed
- IP addresses, cookies, session IDs, page content, form data
Analytics export for AI referral traffic
A periodic export from your analytics tool (GA4, Adobe Analytics, Amplitude) showing aggregated referral traffic from AI assistants to your owned properties. One-way push from your tool to Beacon; we do not query your analytics tool directly.
- Fields received
- Aggregated session counts, referrer domain, landing URL, time window
- Fields not received
- Individual user identifiers, IP addresses, cookie IDs, per-person conversion data
SEO data export
Existing SEO data (keyword search volumes, ranking positions, click-through rates), typically exported from Search Console, Ahrefs, or SEMrush. Aggregate, non-PII. Lets Beacon correlate AI assistant visibility with organic search performance.
Industry benchmarking
An opt-in exchange: you contribute your scoring artifacts (aggregated and anonymized at the category level), and in return you see how your brand compares against the rest of your category inside the dashboard. Participation requires written authorization and can be withdrawn at any time. Withdrawal stops future contributions; previously aggregated, anonymized data remains in historical benchmark series.
Security controls
Encryption
- At rest. AES-256 via AWS-managed keys.
- In transit. TLS 1.2 or higher for all customer-facing endpoints; TLS 1.3 where the client supports it.
Access control
- Dashboard. Authentication via Clerk. Enterprise SSO (SAML / OIDC) available on request. Role-based access via Brand-User membership: Owner, Admin, Editor, Viewer.
- Beacon staff. Scoped access to production via SSO, time-bounded credentials, and least-privilege IAM. Production database access is restricted, audit-logged, and limited to the engineers who require it.
- API access. Scoped, brand-isolated API tokens, revocable from the dashboard.
Audit logging
Dashboard authentication events, brand-level read access, administrative actions, and production database access are logged with timestamp, actor, and target.
Secret management
Application secrets are stored in Vercel and AWS secret managers, never in source control. CI/CD pipelines pull secrets at deploy time. Database credentials and API keys are rotated on staff offboarding.
Network
The production database is private to a VPC and is not internet-reachable. The dashboard and worker tier reach the database over private network only. Egress to sub-processors is over public TLS endpoints; no inbound traffic from sub-processors is required.
Privacy and compliance
Beacon does not process consumer personal data. For the limited dashboard-user account information Beacon holds (your team's Beacon users), Beacon can enter into a Data Processing Agreement and acts as a service provider under documented customer instructions, covering GDPR, CCPA, and CPRA obligations as applicable. Beacon does not sell or share personal information. Your rights over that information, and how the website and the app handle visitors' and users' data, are in our Privacy Policy.
Beacon does not receive or process PHI under the baseline product and is not currently covered by a Business Associate Agreement.
As Beacon scales into the enterprise tier, we are bringing our control environment in line with industry frameworks (SOC 2, ISO 27001). We are happy to walk InfoSec teams through our current control mapping under NDA.
Data Processing Agreement
Where an engagement involves processing personal data, Beacon can put a Data Processing Agreement in place alongside the master service agreement, covering sub-processor disclosure, international transfers under Standard Contractual Clauses where applicable, technical and organizational security measures, customer audit rights, and breach notification within 72 hours of a confirmed incident.
Termination and data export
On termination, Beacon provides a complete export of your account data: the AI answers Beacon collected for you, scoring artifacts, playbooks, and configuration. Customer data is deleted from production systems within 30 days of contract termination or explicit deletion request. If a customer relationship ends without an explicit deletion request, data is automatically deleted within 12 months. Backups are purged on the standard backup rotation. A deletion certificate is available on request.
Questions
Does Beacon need a network integration into our environment to operate?
No. The baseline product requires no integration on your side. You provide your brand name, primary domain, and strategic direction at onboarding; we do the rest.
Does Beacon access our analytics, ad accounts, CRM, or CMS?
No. The optional analytics export is a one-way push from your tool to Beacon, not a connection from Beacon to your tool.
Does Beacon receive our customers' personal data?
No. The baseline product works on public AI assistant answers, not consumer data. Each optional integration is configured so that consumer-identifying fields are excluded at source and are also discarded at Beacon's ingest layer as a defense-in-depth control.
Are our prompts or our customers' AI conversations sent to Beacon?
No. Beacon does not see your customers' conversations with AI assistants. Beacon poses its own buyer-style questions to public AI assistants and reads those answers.
Where is our data stored?
The United States. Database and object storage are hosted on AWS in us-east-1. Application hosting is on US infrastructure.
Are AI models trained on our data?
No. Beacon does not fine-tune or train models on customer data. The model providers Beacon uses operate under API terms that exclude API customer traffic from model training.
Can a business-side user enable an optional integration without security review?
No. Each optional integration is disabled by default and requires written authorization from an authorized customer representative to enable. Customers can also specify in the master agreement which optional integrations are permitted; everything else is blocked at the contract and ingest level.
Does our usage feed industry benchmarks automatically?
No. Benchmark contribution is opt-in and requires written authorization.
What happens to our data if we terminate?
You receive a complete export of your account data. Beacon deletes your data from production within 30 days; backups are purged on the standard backup rotation. A deletion certificate is available on request.
Can we host Beacon in our own environment?
Beacon is delivered as a managed SaaS product. We do not currently offer a self-hosted deployment.
Changes to this page
Beacon may update this page from time to time. Material changes are posted on this page with an updated "Last updated" date. Continued use of Beacon after changes are posted constitutes acceptance of the updated page.
Contact
For security, privacy, legal, and procurement inquiries: privacy@usebeacon.ai.
Beacon AI Technologies, Inc., Los Angeles, CA.