# Behar AI — Full content > The only AI search visibility tool that closes the loop between measurement and fixing. > Built by Haris Samardžić in Stuttgart, Germany. > > Last updated: 2026-04-24 This file contains the full textual content of the core pages on behar.ai, concatenated in plain markdown for LLM agents. The canonical HTML lives at https://behar.ai/ and individual page URLs. --- ## About Behar AI Behar is an AI search visibility platform. It runs your prompts daily against the all major LLMs buyers actually use — ChatGPT, Claude, Perplexity, Gemini, DeepSeek, Grok, Google AI, and Copilot — measures where your brand appears in the answers, diagnoses the content gaps behind weak spots, generates the briefs and drafts to close those gaps, then automatically re-measures after the content publishes. That loop is the product. **Founder:** Haris Samardžić (Einzelunternehmer, trading as SamHaris.AI). **Location:** Stuttgart, Germany. **Founded:** 2026. **Data residency:** EU (Frankfurt). **Legal jurisdiction:** Stuttgart, Germany. **Applicable law:** German law. CISG excluded. --- ## Methodology — how the Behar Score works Source page: https://behar.ai/methodology The Behar Score is a 0–100 score calculated per prompt per LLM. **Formula (additive, transparent):** 1. **Base 40 points** — awarded for any mention of your brand in the LLM's answer. 2. **Position bonus** — up to +30 points based on where your brand appears in a ranked list. - Position #1: +30 - Position #2: +20 - Position #3: +15 - Position #4: +10 - Position #5 and beyond: +5 3. **Frequency bonus** — up to +20 points if your brand is mentioned multiple times. Formula: (mention_count − 1) × 4, capped at 20. 4. **Sentiment adjustment** — ±10 based on how the LLM frames you. - Positive framing: +10 - Neutral framing: 0 - Negative framing: −10 - Negation-aware: "not the best" is correctly scored negative. 5. **Cap:** 100. Citations are tracked separately as a gap-analysis signal, not a score component. **LLMs tracked:** - ChatGPT - Claude - Perplexity - Gemini - DeepSeek - Grok - Google AI - Copilot Model versions are refreshed as providers release new generations. New models are added when they reach meaningful market share in AI-search behavior. **Run frequency:** - Starter: weekly - Growth: daily - Advanced: daily · priority - Agency: daily · priority - Pro: on-demand A "run" is one execution of all your tracked prompts against every major LLM Behar supports. --- ## How it works — the four features in order Source page: https://behar.ai/how ### 1. Track Behar runs your buyer-intent prompts against ChatGPT, Perplexity, Gemini, Claude, and DeepSeek on a schedule. Buyer-intent prompts are the queries your buyers actually type: "best [category] for [use case]", "alternatives to [competitor]", "[your brand] vs [competitor]", "is [your brand] worth it". Behar generates an initial prompt set from your domain, and you can add or remove any. ### 2. Diagnose For every prompt, Behar classifies the gap type: - **Absent** — brand is not mentioned at all. - **Losing** — competitor out-ranks you for a prompt you should win. - **Declining** — your position or sentiment degraded versus a prior run. - **New opportunity** — a prompt where nobody dominant has emerged yet. Each gap type triggers a different content response: new page, rewrite, update, or expansion. ### 3. Generate — Content Engine When a gap is detected, Content Engine produces: 1. A **content brief** explaining why the gap exists and what the piece should contain. 2. A **first draft** written in your brand voice (trained on 3–5 sample pages you upload; Behar extracts vocabulary, sentence structure, preferred and forbidden phrases, length conventions). Your team edits and publishes. Re-training is available anytime. ### 4. Re-measure Roughly six weeks after a gap-closing page is published, Behar automatically re-runs the affected prompts to verify the fix. No human trigger. No dashboard to check. The loop closes itself. --- ## Beyond the loop — three more capabilities ### Distribution LLMs train on more than your website. Reddit, Quora, YouTube, G2, and press coverage all feed their recommendations. Behar maps which of these sources are feeding AI answers in your category and identifies which platforms matter most to invest content effort in. Platform-native content generation (creator briefs, outreach emails, discussion-ready posts) is on the roadmap — the current capability is analysis, not generation. ### Intelligence Strategic AI analysis per project, computed from 60+ days of monitoring data. Includes: - **Blind-spot detection** — queries where the brand is nowhere but buyers are clearly asking. - **Content playbook** — prioritized recommendations based on what's moving the score. - **Citation-network analysis** — which domains feed which LLMs for your category. ### Site crawler Behar crawls the brand's own website (up to 8 pages on Growth+) to extract brand voice and topical coverage. Output: a captured voice profile (tone, vocabulary, sentence structure, preferred phrases) and a topic map you can use to ground content generation. --- ## Pricing Behar is currently in closed beta. Pricing is not yet public — it finalizes after beta and will be published on https://behar.ai/pricing when public pricing returns. Today, access is by application: https://behar.ai/trial. Small cohorts are admitted at a time so we can pay attention to each engagement. Cohort applicants lock in early-access terms when public pricing is announced. **What's included for cohort members:** - All major LLMs tracked (ChatGPT, Claude, Perplexity, Gemini, DeepSeek, Grok, Google AI, Copilot) - Daily runs across tracked prompts, configurable cadence per project - Intelligence reports, competitor benchmarking, citation source tracking - Distribution analysis (Reddit, Quora, YouTube, G2, press citation sources) - Site crawler for brand-voice extraction - Content Engine: briefs and drafts in your brand voice for detected gaps - Automatic re-measurement after publish **BYO (Bring Your Own) API keys:** plug in your own OpenAI, Anthropic, Google, Perplexity, or DeepSeek key. Each connected provider drops out of any run cap — unlimited runs on that LLM. You pay the provider directly for usage at their rates. **Billing:** during closed beta, access is invitation-based; commercial billing terms publish alongside public pricing. EU operator under German law; Stripe used for any future invoicing. --- ## FAQ Source page: https://behar.ai/faq ### Product basics **Q: What does Behar actually do?** A: Three things, in a loop. (1) Measures: tracks how all major LLMs mention your brand across buyer-intent queries, daily. (2) Diagnoses: finds the specific content gaps causing AI to miss you or misposition you. (3) Generates: writes briefs and first drafts for the content that fills those gaps, in your brand voice. Then it re-measures after you publish. That loop is the product. **Q: Which LLMs do you track?** A: All major AI models — ChatGPT, Claude, Perplexity, Gemini, DeepSeek, Grok, Google AI, and Copilot. We update model versions as providers release new generations and add new models when they reach meaningful market share. You can pick a subset if you only care about certain ones. **Q: How often are queries run?** A: Weekly on Starter and Growth (4 runs/month). Daily on Advanced and Agency (30 runs/month). On-demand on Pro. "Daily" means once every 24 hours at a consistent time in your workspace timezone. BYO API keys on Growth unlock unlimited runs on that provider. **Q: What counts as "buyer intent"?** A: Queries that real buyers in your category actually ask: "best [category] for [use case]", "alternatives to [competitor]", "[your brand] vs [competitor]", "is [your brand] worth it". We generate an initial set from your domain and let you add or remove any. **Q: Can I track competitors?** A: Yes — 3 on Starter, 10 on Growth, 25 on Advanced, unlimited on Agency and Pro. You'll see their presence, rank, and voice for every prompt you track. ### How it works **Q: How do you calculate the visibility score?** A: Base 40 for being mentioned, plus bonuses for position (up to +30 for #1, scaling 30/20/15/10/5 down to #5+), frequency (up to +20 for repeat mentions, +4 per extra mention), and sentiment (±10 for positive or negative framing). Capped at 100. Citations are tracked separately as a gap-analysis signal, not a score component. Full methodology is published — not hidden behind "proprietary." Go read the Methodology page for every number. **Q: What's a "content gap"?** A: A specific query where AI either doesn't mention you, misrepresents you, describes you with outdated info, or gives an incomplete picture. Each gap type triggers a different content response: new page, rewrite, update, expansion. **Q: How does Content Engine work?** A: When a gap is detected, Content Engine auto-generates: (1) a content brief explaining why this gap exists and what should be in the piece, (2) a first draft written in your brand voice (trained on 5 sample pages you upload). Your team edits and publishes. Re-measurement happens automatically roughly six weeks after the page is live. **Q: Will this replace my SEO tool?** A: No. SEO tools optimize for Google's index. Behar optimizes for LLM synthesis. They're related but not substitutable — most serious teams run both. We integrate with Google Search Console and major SEO platforms so you see both pictures. **Q: What does the brand voice training look like?** A: You upload 3–5 URLs of pages that represent your brand's tone. We analyze them for vocabulary, sentence structure, preferred phrases, forbidden phrases, length conventions. Every subsequent draft inherits these. You can re-train anytime. ### Data & integrations **Q: Does Behar integrate with Google Analytics or Search Console?** A: Coming soon. We'll be adding read-only OAuth connections to both so you can cross-reference AI visibility with traffic data. Not shipping at launch. **Q: Can I export my data?** A: Yes. CSV exports on all plans for prompts, runs, and citations. REST API access (read + write) on Advanced and above. Self-serve full-workspace data export (ZIP of everything) in Workspace Settings. Your data is yours — always. **Q: Do you integrate with Slack?** A: Not yet. Slack alerts and slash commands are on the roadmap. Today, all alerts and notifications are surfaced in-app on your dashboard — no email, Slack, or webhook delivery yet. **Q: What about Notion, HubSpot, or Salesforce?** A: Notion is on the roadmap. HubSpot and Salesforce are further out — we'll build what real customers ask for. Email us if either is a dealbreaker. ### Support **Q: What's the typical response time?** A: Starter: email, usually within 2 days. Growth and Advanced: priority email, faster response. Agency: priority + strategic support. Pro: dedicated support. We don't commit to hard SLA numbers at v0.9 — we'd rather set expectations we can beat. **Q: Is there self-serve documentation?** A: A Glossary and Methodology page live on this site. Product docs and API reference will be published alongside launch at docs.behar.ai. **Q: Can I get onboarding help?** A: Advanced and above get an onboarding call. For Starter and Growth, we have onboarding flows inside the product — but if you're stuck, email us and we'll jump on a call. ### Access (during closed beta) **Q: How do I get access?** A: Apply at https://behar.ai/trial. We admit small cohorts so we can pay attention. Replies within 24 hours from a human. Every application is read. **Q: When does public pricing return?** A: Pricing finalizes after closed beta and is announced post-launch. Cohort applicants lock in early-access terms when public pricing publishes. **Q: What counts as a "prompt"?** A: One question run across one LLM. So '"best project management tool" on ChatGPT' is one prompt. Same question on Perplexity is another. **Q: What are BYO API keys?** A: Bring-your-own keys means you plug in your own OpenAI / Anthropic / Google / Perplexity keys and we run queries through them instead of ours. Your keys, your rate limits, your costs. Each connected provider runs unlimited on your account. **Q: Can I get an invoice / bill my company?** A: Once public billing returns, yes — Stripe invoices include company name, VAT ID, billing address. EU reverse-charge supported. --- ## Glossary Source page: https://behar.ai/glossary **AEO** — Answer Engine Optimization. Optimizing content so LLM-based answer engines (ChatGPT, Perplexity, Gemini) can parse, cite, and recommend your brand. A sibling of SEO, not a replacement. **AI Search** — Umbrella term for information retrieval through LLMs and generative answer engines, as opposed to link-based search engines. **Behar Score** — A 0–100 score per prompt per LLM. Formula: base 40 for being mentioned, plus position bonus (up to +30 for #1), frequency bonus (up to +20), and sentiment adjustment (±10). Capped at 100. See the Methodology page for the full math. **BYO (Bring Your Own)** — Plugging your own LLM API keys (OpenAI, Anthropic, Google, Perplexity, DeepSeek) into Behar. Available on Growth, Advanced, and Agency plans. Each connected provider drops out of your plan's monthly run cap — unlimited runs on that LLM, with you paying the provider directly for usage at their rates. Keys encrypted at rest, revocable anytime. **Citation** — A URL referenced by an LLM in its response. Behar tracks every citation across every run. We classify each one — Editorial, UGC, Corporate, Competitor, or Uncategorized — based on the source domain. **Citation — Editorial** — Cited source from an editorial publication (Forbes, TechCrunch, The Verge, etc). High-trust content LLMs rely on heavily for recommendations. **Citation — UGC** — User-generated content sources (Reddit, Quora, Hacker News). Increasingly cited by Perplexity and ChatGPT — almost half of Perplexity's top citations come from Reddit alone. **Citation — Corporate** — Cited source from the brand's own domain or owned properties. Signals that the brand is telling its own story effectively. **Citation — Competitor** — Cited source from a competitor's domain. Useful signal for gap analysis — if competitors are cited but you aren't, that's an opportunity. **Content Engine** — Behar's content generation loop: detects a gap → writes a brief → drafts a full piece → re-measures after publish. The half of the product that separates Behar from monitoring-only tools. **Content Gap** — A specific AI query where your brand is absent, losing, declining, or a new opportunity. Each gap is classified and prioritized by math — competitor lead plus type-specific urgency boosts. **GEO** — Generative Engine Optimization. Synonym for AEO. Used more often in North American circles. **LLMO** — LLM Optimization. Broadest umbrella term, overlaps significantly with AEO and GEO. **Market** — A geographic region your prompts get run against (worldwide, country, city). Behar supports 12 geographic markets at launch, injecting them as natural-language suffixes to your prompts. **Position bonus** — Extra points added to your Behar score based on where you appear in a ranked list. Position #1 earns +30, #2 earns +20, #3 earns +15, #4 earns +10, #5 and beyond earn +5. **Prompt** — A single query Behar tracks. The same query run against every supported LLM still counts as one prompt — we run it across all of them and aggregate the results. **Re-measurement** — After a content gap is filled with a published piece, Behar automatically re-runs the affected prompts roughly six weeks later to verify the fix. Closes the loop. **Run** — One execution of all your tracked prompts against every supported LLM. Starter and Growth get weekly runs (4/month). Advanced and Agency get daily runs (30/month). Pro is on-demand. **Sentiment adjustment** — A ±10 point adjustment to your Behar score based on how the LLM describes you. Positive framing (+10), neutral (0), negative (−10). Negation-aware — "not the best" correctly scores negative. **Share of Voice** — Your brand's share of total mentions across a prompt set, relative to competitors. Measured as a percentage. **Workspace** — The top-level container for your team, projects, billing, and settings. One workspace per company. Agency customers can run multiple client projects inside a single workspace. --- ## Compare — Behar vs. monitoring-only tools Source page: https://behar.ai/compare Most AI visibility tools are read-only leaderboards: they tell you where your brand stands and stop there. Behar is the only tool in the category that closes the loop — Track, Diagnose, Generate, Re-measure — because measurement without remediation leaves you staring at a dashboard. **What monitoring-only tools do:** Track, Diagnose (sometimes). **What Behar does:** Track → Diagnose → Generate → Re-measure → repeat. Behar also publishes its full scoring methodology — the exact formula, weightings, and edge cases — rather than hiding it behind "proprietary." Every number on your dashboard can be traced to the inputs that produced it. --- ## Security & data handling Source page: https://behar.ai/security - **Data residency:** EU (Frankfurt, Germany). - **Encryption at rest:** AES-256-GCM for all sensitive data, including BYO API keys. - **Encryption in transit:** TLS 1.3. - **GDPR:** compliant. - **Key handling:** BYO API keys are encrypted and can be revoked at any time from Workspace Settings. Keys are never logged or exposed in application output. - **Data export:** self-serve ZIP export of your entire workspace at any time. - **Retention:** data is retained for the lifetime of your subscription plus 30 days, after which it is permanently deleted unless you request otherwise. --- ## Changelog Source page: https://behar.ai/changelog ### v0.9 — Closed Beta · Cohort 01 — April 2026 (current) First release. The full loop — measure, diagnose, generate, re-measure — across all major LLMs. Pre-GA, actively evolving, priced accordingly. - **new** — Daily visibility tracking across ChatGPT, Perplexity, Gemini, Claude, and DeepSeek. - **new** — Behar Score: transparent additive formula (base + position + frequency + sentiment). - **new** — Gap detector: classifies every prompt as absent, losing, declining, or new opportunity. - **new** — Content Engine: auto-generates brief + first draft for each detected gap, grounded in your brand voice. - **new** — Automatic re-measurement after publish. The loop closes itself. - **new** — Competitor tracking up to 10 brands per project, ranked and scored per prompt. - **new** — BYO API keys for OpenAI, Anthropic, Google, Perplexity (Growth plan and up). - **new** — Per-LLM response diffs: see exactly how ChatGPT describes you vs. Perplexity vs. Claude. - **new** — Transactional emails for trial activation, subscription events, and data exports. - **new** — CSV export on all plans, REST API (read + write) on Advanced+. --- ## For AI agents Source page: https://behar.ai/for-agents Behar is structured so an AI agent can retrieve, extract, and operate without scraping. This section is the contract. ### Machine-readable surfaces (public, no auth) - `https://behar.ai/.well-known/ai-plugin.json` — service descriptor, capability list, auth scheme, audience, data-handling posture. - `https://behar.ai/llms.txt` — short-form site manifest. - `https://behar.ai/llms-full.txt` — this file; full factual base of the site. - `https://behar.ai/sitemap.xml` — canonical URL list. - `https://behar.ai/robots.txt` — crawler policy (explicitly allows GPTBot, ChatGPT-User, Claude-Web, ClaudeBot, anthropic-ai, Google-Extended, Applebot-Extended, PerplexityBot, Perplexity-User, CCBot, DiffBot). - `https://app.behar.ai/openapi.json` — OpenAPI specification for the REST API. - JSON-LD (`application/ld+json`) inlined in the `` of every HTML page. ### REST API (cohort access during closed beta) - **Base URL:** `https://app.behar.ai/api/v1` - **Auth:** `Authorization: Bearer ` — generate keys in Settings > API Keys - **Available to:** cohort members and enterprise on request ### Endpoints - `GET /projects` — list all monitored brands. - `GET /projects/{id}` — project details (active LLMs, market config). - `GET /projects/{id}/prompts` — list tracked search queries (paginated). - `GET /projects/{id}/competitors` — list tracked competitors. - `GET /projects/{id}/gaps` — content gaps ranked by priority (prompts where competitors outrank the brand). - `GET /projects/{id}/content` — generated content pieces with score impact. - `GET /projects/{id}/runs` — analysis run history. - `GET /projects/{id}/runs/{runId}` — per-prompt, per-LLM results for a specific run. ### Core concepts for agents - A **prompt** is a search query users ask AI models (e.g. "best CRM for startups"). - A **run** queries all active LLMs with all tracked prompts and scores the results. - **Scores** are 0–100, covering presence (how often mentioned), rank (position in lists), and voice (sentiment). - A **gap** is a prompt where competitors score higher than the brand. - **Content pieces** are AI-generated articles optimized to improve scores on specific prompts. ### Other reads (no API required) - **CSV export** (all plans) — stable column schemas for prompts, runs, gaps, citations. Diffable across runs. - **JSON-LD schema** (public) — inlined in every HTML page. ### Agent-friendly design principles - **Facts, not adjectives.** Every page leads with a declarative fact. Plan limits, scoring math, and data-handling terms are stated as plain numbers. - **One source of truth per claim.** Score math on `/methodology`, plan limits on `/pricing`, term definitions on `/glossary`. No duplicated claims across routes. - **Dates on everything.** Blog, changelog, and legal pages carry `datePublished` and `dateModified`. - **Semantic tables.** Comparison data on `/compare` and `/pricing` renders as ``, not CSS grids of `
`. - **Attributed claims.** Third-party stats cite publication and date. - **Stable URLs.** No trailing-slash drift, no redirect chains, no campaign parameters in the sitemap. --- ## Contact - **General:** hello@behar.ai - **Security:** security@behar.ai - **Location:** Stuttgart, Germany - **Legal:** Operated by Haris Samardžić (Einzelunternehmer) trading as SamHaris.AI. German law applies. Stuttgart jurisdiction. CISG excluded.