MentionHub - AI SEO Visibility & Brand Tracker

Your SEO report no longer explains what is happening

Rankings hold. Traffic falls. The reason is that buyers stopped clicking ten blue links and started asking ChatGPT, Gemini and Perplexity a question — and those tools answer with one narrative answer. If your brand is not named in it, you are not there. There is no page two to fight for.

MentionHub measures that. It asks the questions your customers actually ask, over and over, across several AI engines, and reports how often each engine names your brand versus your competitors — as a rate with a confidence interval, not as a single anecdote.

Why it is a measurement, not a screenshot

Language models are not deterministic. Ask the same question three times and you get three answers. A tool that fires one query and tells you "you rank #2" is reporting a coin flip.

MentionHub takes N samples per prompt per engine (3-10, default 5), then reports mention rate, share of voice, average position and citation share, each with a 95% Wilson confidence interval. A change between two periods is flagged only when a two-proportion z-test says it is real — not when the number simply looks different.

Who buys this

• SEO and marketing agencies — one installation, one workspace per client, white-label PDF reports with your own logo and no trace of MentionHub.
• SaaS and e-commerce owners — see whether the engines recommend you or your competitor when a buyer asks for "the best X for Y".
• Anyone reselling capacity — connect the provider keys once, set your own credit rate, and sell sampling to tenants as credits.

What it does not do

Stated up front, because an honest boundary is worth more than a surprise after purchase.

• It reads the provider APIs, not the consumer ChatGPT, Gemini or Perplexity apps. Related surfaces, not identical ones.
• It does not produce exact rankings. Every figure is a statistical estimate shown with its interval.
• It does not scrape Google or any other search engine.
• It does not write content for you.
• Sentiment is a local lexicon over the sentence around each mention, not a model reading the whole reply.

Built with

Laravel 12, Vue 3 with Inertia 2, MySQL, Tailwind CSS. Full source code, a complete offline HTML manual covering cPanel and VPS installation, the cost model, agency mode and troubleshooting, plus a style guide for rebranding colours, logo and product name.

Live demo

Customer workspace login: demo@mentionhub.io / demo1234 The demo resets daily and is seeded with realistic sampling data. Admin panel credentials are included in the item documentation.

Changelog

Version 1.0.0 — Initial release.

Features

• Brands and competitors — unlimited brands per organization, up to 10 tracked competitors each, alias matching (Acme, Acme Inc, acme.com) with word-boundary and accent-folded detection, plus competitor auto-suggestion from what the engines actually say.
• Prompts without the blank page — 8 industry packs with 128 curated questions (SaaS, e-commerce, local services, agency, clinic, food and drink, education, real estate), an AI generator that writes prompts from the brand profile, CSV import, and manual entry. Four intents: discovery, comparison, problem, branded.
• Three engines — OpenAI (ChatGPT), Google Gemini and Perplexity, each with a model catalogue and its real per-million-token rate. Adapters sit behind an interface, so a fourth engine touches nothing else.
• Analytics — mention rate with delta, share of voice, average position, citation share (being linked is not the same as being named), win rate head to head, competitor heat map, and per-engine comparison. Being strong on one engine and absent on another is itself the finding.
• Read the raw answers — every stored sample, with the mention highlighted. This is the evidence an agency shows its client.
• Trends and annotations — daily, weekly and monthly history, statistically significant change detection, and event markers so a client can see cause and effect.
• Site AI-readiness audit — llms.txt presence and validity, robots.txt rules for GPTBot, PerplexityBot, ClaudeBot, Google-Extended and CCBot, schema.org coverage, Common Crawl indexing, on-page content signals, and a prioritised 0-100 score.
• White-label PDF reports — agency logo, colours and contact details, no trace of MentionHub. Scheduled to multiple recipients on the day and hour you pick, with an option to skip a period in which nothing moved beyond sampling noise. Tokenised read-only share links, delivery archive, and CSV export on every table.
• Cost control — a live estimator shows the monthly cost before a configuration is saved, a hard monthly cap stops runs rather than warning about them, and a cost ledger records spend per run, per engine, per prompt.
• Sell it on — the operator installs the provider keys once, sets a credit rate, and resells capacity to tenants. Credits and provider cost are tracked separately, so the margin is visible instead of guessed at.
• Multi-tenant and agency mode — organizations own the data, roles are owner / admin / member with per-brand limits, quotas are claimed inside a locked transaction (no check-then-write race), and client workspaces can be read-only.
• Super admin panel — engine connection status, tenant list with plan changes, suspend and reactivate (nothing is deleted), manual credit adjustments with a mandatory reason, provider keys and model selection, credit pricing, currency, SMTP with a test send, bank account master data, and every subscription and credit sale in one filterable list.
• Billing — Stripe, PayPal and manual bank transfer end to end: the customer picks an account, uploads proof, and the operator approves inside the transaction detail. Proof is stored outside the web root, prices are read server side, and approval runs through the same replay-safe path as the webhooks. Automatic invoices and billing history included.
• Eight languages, RTL included — English, Arabic, German, Spanish, French, Indonesian, Brazilian Portuguese and Turkish, complete rather than partial, switchable on every page including sign-in and the admin area. Arabic mirrors the layout through CSS logical properties.
• Ready to run — four-screen web installer that seals itself when it finishes (no terminal needed, built for cPanel), demo mode with realistic seeded data and daily reset, GDPR kit with self-service export and deletion, cookie consent, support tickets, and light and dark themes.

Requirements

• PHP 8.2 or higher (8.3 recommended)
• MySQL 8.0 recommended, 5.7.8 minimum — MariaDB 10.4+ also works
• Apache or Nginx with the document root pointed at public/
• One cron entry running every minute
• PHP extensions: bcmath, ctype, curl, dom, fileinfo, filter, hash, mbstring, openssl, pcre, pdo_mysql, session, tokenizer, xml, zip, plus gd or imagick
• An API key from at least one of OpenAI, Google Gemini or Perplexity
• Composer dependencies are already bundled. Node.js is only needed if you want to rebuild the front-end assets — they ship pre-built.

Instructions

1. Upload. Extract the download, then upload the contents of Main Files to your server. On cPanel, upload mentionhub.zip to your account and extract it there.
2. Point the document root at public/. On cPanel this is done in Domains → Document Root. Never serve the application root directly.
3. Create a MySQL database and a user with full privileges on it. Keep the name, user and password at hand.
4. Open https://yourdomain.com/install in a browser. The four-screen wizard checks the server requirements, takes the database credentials, creates the first administrator, and runs the migrations for you. No terminal is required. The installer seals itself when it finishes and returns 404 afterwards.
5. Add one cron entry so scheduled runs, reports and emails fire: * * * * * cd /path/to/your/app && php artisan schedule:run >> /dev/null 2>&1
6. Connect an engine. Sign in as the administrator, open /admin/settings → AI provider keys, paste a key from OpenAI, Gemini or Perplexity, pick a model and press Test. Until at least one engine is connected, no run can start.
7. Set your credit pricing on the same page: tokens per credit, price per credit and the search-call charge. This is where your margin is defined.
8. Configure outgoing mail and payments. SMTP has a test send. Enable only the gateways you actually use — Stripe, PayPal, manual bank transfer, or any combination.
9. Start tracking. Create a brand, add its competitors, pick prompts from an industry pack or write your own, then press Run now. Results appear as soon as the samples come back.
10. Rebrand it. Documentation/style-guide.html in the download explains how to change the colours, logo and product name.

The full manual — cPanel and VPS installation, the cost model, agency mode, troubleshooting — is in Documentation/index.html, offline and in three PDFs.