Measure, diagnose, fix & prove your AI search visibility — in one loop.
BuyerCite polls six AI answer engines on the prompts your buyers actually ask, turns every missed citation into a ranked task, writes claim-gated content grounded in your own evidence, and publishes it to your CMS with schema and /llms.txt — then attributes the AI referral traffic it earns.
-
01 · MeasureRadar & Fan-Out6 engines · Share of Voice · prompt fan-out
-
02 · DiagnoseCitation & Keyword GapsCited-Not-Named · Niche Gate · Cannibalization Guard
-
03 · FixStudio, Catalog & /llms.txt6-stage claim-gated writer · PDP rewriter
-
04 · ProveConnectors & AttributionWordPress Pull · webhooks · zero-IP pixel
6-engine AI answer polling, on your cadence.
Each tracked prompt runs against every enabled engine for the country and language you target. BuyerCite stores the full answer, every cited URL and every brand it names — yours and your competitors' — and rolls them into Share of Voice, citation rate and rank trends on the Executive Pulse.
- Cadence you control: Daily (31 runs per period), 3× per week (14) or Weekly (5).
- Engine health checks: a fixed grounded probe opens a circuit when an engine degrades, so a bad run is deferred instead of recorded as "not cited".
- Fair metering: credits are reserved before a poll and released automatically when an engine fails.
- Prompt fan-out discovery (6 Radar Credits): expands one head query into the sub-questions AI engines silently research, so you track what buyers really ask.
Live web-grounded answers with cited sources.
The generative summary above classic results.
Google's conversational, fan-out search mode.
Answer engine with dense inline citations.
Gemini answers grounded with Google Search.
Claude answers with web search citations.
Catch every CITED_NOT_NAMED answer.
The most expensive gap in AI search is invisible in rank trackers: the engine links to your page as a source but recommends a competitor — or calls you by a legal entity name buyers have never heard of. BuyerCite resolves your brand aliases (trading names, product lines, domain variants) during onboarding and checks every captured answer against them.
Prompt "best project management software for a remote agency" → the answer cites example.com/features/project-management but names two competitors. BuyerCite opens a gap task with the cited URL, the competing brands and a counter-guide brief.
24 AI crawlers, policy vs. reality.
The GEO audit separates what your robots.txt says from what your server actually returns to each AI user agent — checked only against your own verified domain, labelled "simulated", never by impersonating a crawler to fetch anyone else's site.
- GPTBot
- OAI-SearchBot
- ChatGPT-User
- Google-Extended
- ClaudeBot
- Claude-User
- Claude-SearchBot
- Meta-ExternalAgent
- PerplexityBot
- Applebot-Extended
- GoogleOther
- GrokBot
- anthropic-ai
- Bytespider
- CCBot
- cohere-ai
- Diffbot
- FacebookBot
- ImagesiftBot
- Meta-ExternalFetcher
- Omgilibot
- Bingbot
- Googlebot
- Perplexity-User
Honest scoring: WAF-blocked or empty client-rendered pages are reported as "not measurable" and readiness is capped at 70 — never a fabricated pass.
One ranked queue of fixes — on-niche and non-cannibalizing.
Every missed citation, Cited-Not-Named answer and fan-out sub-question becomes a scored gap. Two deterministic guards keep the queue honest before anything reaches your calendar.
Semantic relevance before scoring
Candidate keywords are embedded and compared with your Brand DNA using pgvector similarity. Junk intent (jobs, coupons, logins), competitor navigational queries and off-niche topics are rejected with a stated reason, before any score is shown.
No page competes with itself
Before a new guide is queued, BuyerCite checks it against your existing URLs and planned articles. Overlaps are routed to "improve the existing page" instead of producing a near-duplicate that splits citations.
Prioritized, scheduled, owned
Gaps are ranked by commercial value and winnability, then dragged onto a 30-day content calendar. Off-site citation tasks (directories, reviews, comparison pages) are tracked alongside on-site work.
A 6-stage writer that refuses to hallucinate.
Generic AI writers invent specs, warranties and prices. BuyerCite's buying guides are compiled, not improvised: every factual claim is bound to evidence, and your Compiled Brand Rules (tone, banned phrases, compliance language) are enforced deterministically.
-
Stage 1Evidence Ledger
A per-article dossier of facts from your site, catalog and approved documents — each with its source.
-
Stage 2Rule-checked outline
An H2 plan from the template's mandatory blocks, with facts allocated to sections.
-
Stage 3 / 3bDraft + zero-tolerance Claim Gate
Citation-ready draft; every price, spec, warranty and certification claim must match the ledger or the block is blocked.
-
Stage 4Citation-readiness evaluator
6 dimensions and 18 sub-dimensions scored, with a bounded rewrite loop for weak sections.
-
Stage 5Visuals
On-brand images with descriptive alt text, generated or regenerated on demand.
-
Stage 6Compile + publish gates
Sanitized canonical HTML, validated JSON-LD and a final claim sweep before anything can publish.
Human-in-the-loop by default: unsupported claims stop publishing until a reviewer resolves them. A grounded article costs 10 Studio Credits.
Make your catalog machine-readable.
AI shopping answers are built from structured facts. BuyerCite turns your product data into the formats answer engines trust.
- Catalog CSV import & sync. Bring products, specs, prices and availability in once; they become the evidence the writer and rewriter are allowed to use.
- Product & Offer JSON-LD. Valid schema compiled from the catalog row — never a placeholder value.
- Grounded PDP rewriter. Spec tables and FAQs rewritten for extractability, with a numeric guard that tokenizes every quantity with its unit (3/4 HP, 3000W, ½ HP) so 500 kW can never pass for 500 W. 1 Studio Credit live, or half a credit per product in batch.
A curated Markdown map of your brand facts, best guides and key products, served from your own domain via the WordPress plugin or the Content Pull API (/content/v1/sites/{site_key}/llms.txt).
# Acme Corp > Project management software for distributed > teams. Serving the US, UK & Europe. ## Buying guides - [Choosing a project management tool](https://example.com/guides/choosing) ## Products - [Acme Cloud](https://example.com/product/cloud): team workspaces
See ours live: buyercite.com/llms.txt
Publish where your site already lives.
Approved content leaves BuyerCite through authenticated, auditable channels. Nothing is pushed with stored admin passwords.
Your site pulls approved guides on WP-Cron using a scoped bcpull_… token. BuyerCite stores only a SHA-256 hash for lookup plus an AES-256-GCM envelope so Owners and Admins can reveal it in an audited action. The first handshake binds the token to your site host.
The plugin also serves /llms.txt and forwards AI bot logs.
Receive article.published, article.updated and article.unpublished events on any HTTPS endpoint. Every delivery is signed with HMAC-SHA256 over timestamp + "." + body using your bcwhsec_… secret, and failed deliveries are retried on a fixed schedule.
POST /your-endpoint HTTP/1.1 X-BuyerCite-Event: article.published X-BuyerCite-Timestamp: 1795000000 X-BuyerCite-Delivery-Id: 7f3c2a90-… X-BuyerCite-Signature: sha256=<hex HMAC(secret, timestamp + "." + body)> Content-Type: application/json
Headless sites can poll the Content Pull API instead (GET /content/v1/sites/{site_key}/changes with an opaque cursor and ETags).
Public REST API.
Read Radar overviews, gaps, keywords, articles and catalog products; enqueue gap articles, fan-outs and GEO audits. Authenticate with an organization API key — bc_sk_live_… for production or bc_sk_test_… for testing — stored only as a hash.
curl https://buyercite.com/api/v1/workspaces \ -H "Authorization: Bearer bc_sk_live_…" HTTP/1.1 200 OK X-BuyerCite-RateLimit-Limit: … X-BuyerCite-RateLimit-Remaining: … X-BuyerCite-Radar-Credits-Remaining: … X-BuyerCite-Studio-Credits-Remaining: …
API call allowances are included on Growth and Scale. See pricing.
Stateless MCP endpoint.
Connect your own AI agent to BuyerCite over JSON-RPC at POST /mcp. Each tool is scoped to the key's permissions, and long jobs return a handle you poll with get_job.
- list_visibility_gapsOpen gaps by type and score
- expand_query_fanoutDiscover sub-questions
- audit_url_geoRun a GEO page audit
- enqueue_gap_articleQueue a claim-gated guide
- sync_llms_txtRebuild /llms.txt
- get_jobPoll a running job
See the traffic and revenue AI answers send you.
Zero-IP first-party pixel
A lightweight script on your domain records page views and conversions with an ephemeral first-party session id. It stores zero IP addresses and sets no third-party cookies.
AI referral channels
Referrers and utm_source values from ChatGPT, Perplexity, Gemini, Claude and others are classified as ai_referral with the engine attached, next to organic, direct, paid and social.
Crawler logs & Search Console
See which AI crawlers fetch which pages. Client IPs in bot logs are transient and nulled within 24 hours.
Run the whole loop on your own domain.
Start with 150 Radar Credits and 20 Studio Credits for 7 days — or see what AI engines see today with a free audit.