Tanitio is an AI-assisted marketing operations platform that runs social media, advertising, measurement, SEO/GEO and e-commerce marketing from a single panel. It was designed for agencies, online stores, content creators and local businesses. Every brand works in its own isolated workspace, so content, calendars, ad accounts, measurement connections and AI memory never mix between brands.

Context and need

A brand's marketing is typically spread across a dozen separate tools: Meta Business Suite, Google Ads, GA4, Search Console, Tag Manager, SEO crawlers, competitor trackers and store dashboards. Each has its own login and its own reports, and the team stitches the data together by hand every week. At the same time, many AI tools change campaigns without any oversight. Tanitio started from these two problems: bring the scattered operation into one screen, and turn AI into an assistant that only acts with human approval. A metric that cannot be measured is never estimated; it is reported as “not measured”.

Solution and scope

Content and publishing

The content calendar, weekly plan generation, recurring schedules, templates and hashtag groups work in one flow. Tanitio connects to Facebook, Instagram (feed, Reels, stories), X, LinkedIn, YouTube and Telegram to publish directly and measure results; for TikTok it sends content to the creator's draft inbox. A content guard runs at publish time and flags risky claims, placeholder links and wrong-language text. Server-rendered card templates produce images in square, portrait and story formats, and raw footage is turned into subtitled vertical Reels automatically with Whisper and ffmpeg.

Advertising and measurement

Google Ads and Meta Ads accounts are read down to campaign, ad group, search term and creative level, and a unified ads overview shows spend, conversions, the GA4 funnel and a profit estimate side by side. New Meta campaigns are always created paused, and targeting is read back to verify it. Meta Pixel and Conversions API diagnostics, Merchant Center feed health, Tag Manager fix suggestions with draft publishing and rollback, GA4, Search Console, Bing and Yandex Webmaster data all live in the same workspace. Measurement health checks catch issues such as wrong currency, wrong time zone and double-counted conversions.

SEO, GEO and competitor intelligence

The site audit covers more than 30 check groups, Lighthouse and CrUX data, content quality and a draft llms.txt; visitors can also run a free SEO scan. The GEO observation module tracks how visible the brand is in AI search, using defined question sets. Competitor discovery finds competitors from search results, and competitor monitoring reports price, stock, CTA and copy changes on home, pricing, product and campaign pages, each with a severity level. Google Trends opportunities are compared against the brand's own Search Console queries.

Controlled AI

The AI produces content packs (headline, description, ad copy, video script, visual idea), scores creatives from 0 to 100, generates recommendations as an ad analyst and updates a weekly brand memory. None of this goes live directly. Changes become a change set: draft, validation and dry run, explicit approval, apply. If the target or configuration changes after validation, apply is rejected and re-validation is required. The outcome of each applied recommendation is measured and fed back into memory. Autonomous optimisation is off by default; when enabled, it runs within percentage and amount caps, a daily action limit, quiet hours, an automatic revert scan every six hours and a per-brand kill switch.

E-commerce and leads

Orders from Shopify, WooCommerce and İkas are synchronised, and product and category SEO is evaluated against a 100-point rubric. Meta lead forms, trackable short links and hosted forms feed a single lead pool; short-link clicks never store raw IP addresses, only a hash with a daily salt.

Management and teams

Access is controlled in three layers: scope (site, social, ads, measurement, e-commerce, email), purchased modules and roles. There are superadmin, brand admin, editor and read-only client portal roles, so agencies can give clients a reporting view without edit rights. A “today's tasks” engine turns recommendations, alerts and findings into tasks automatically. Threshold alerts go to the in-app bell, Telegram and email, and monthly PDF client reports are sent automatically. The superadmin follows actions across all tenants in a single activity ledger.

Technical architecture

The platform is a monorepo built with Bun workspaces. The backend runs on Fastify 5, TypeScript, Drizzle ORM and MySQL 8, with SEO crawl and video jobs in BullMQ and Redis queues. The panel and the public site are one Next.js 16 application with React 19, Tailwind CSS v4, Shadcn/Radix and React Query. The backend has more than 50 modules and the panel more than 80 pages; over 50 scheduled jobs run, from the publisher every five minutes and analytics every six hours to monthly reports. Failures in critical jobs are recorded and reported to the superadmin. Text generation runs through a provider chain of Claude, OpenAI and Groq, and every operation is metered in credits. The interface is in Turkish. The database schema is kept in SQL seed files as the single source of truth.

Security and compliance

• Tenant secrets and platform tokens are stored in AES-256-GCM encrypted columns; there are no default values for secrets.
• HMAC and timing-safe signature checks on webhooks; SSRF protection on crawl endpoints.
• Rate limiting by real client IP, a tenant-scope guard and tenant isolation tests in continuous integration.
• A GDPR/KVKK erasure flow in which preview and execution use the same plan and the database deletion runs in a single transaction.

Outcome and today

Tanitio is live at tanitio.com and offers a two-hour, read-only demo session with a fictional sample brand, no sign-up required. Trial, starter, professional, agency and managed plans run with Stripe, iyzico and manual payment. The open roadmap includes Pinterest and Snapchat publishing, pre-publication content performance prediction and a multilingual interface; these are currently at the planning stage only.