Tolstoy / GEO content platform
Content engineered to be cited by AI.
Tolstoy turns a short content brief into a deeply-researched, fact-grounded, publication-ready blog post built to be quoted by AI answer engines. Describe your brand once, then plan a month of content or request one-off posts.
What Tolstoy is
Grounded, citable content
Every factual claim is traced to a real source and verified verbatim before a post can publish. The output is purpose-built to be quoted by AI answer engines.
Hard cost control
Every workspace has mandatory spend ceilings enforced at the database layer before any AI call is made. The system physically cannot overspend a limit.
Secure multi-tenancy
Strict per-customer data isolation at the database level, invite-only access, and mandatory two-factor authentication for platform operators.
The pipeline
From a one-line brief to a publication-ready post.
A brief flows through four metered AI stages with two human review gates, producing a finished post with full source provenance.
- 01Research
- 02Outline
- 03You approve
- 04Draft
- 05Quality edit
- 06You publish
Research
Three operations run in parallel: Brave web search, a Perplexity Sonar Pro deep-research report, and AI curation. Sources are tiered toward government, academic, and tier-1, every finding is captured with a verbatim quote, and the reader's likely follow-up questions are predicted (query fan-out).
Outline
A binding content plan before any prose: title options, 3 to 10 sections with purpose and word budget, fit-scored product placements, an FAQ spec, comparison tables, and atomic-answer requirements. A malformed outline is blocked before drafting.
Draft
A complete Markdown post: a 100 to 150 word Quick Answer, a TL;DR, body sections, FAQ, and a Sources block, with a claim-provenance map tagging every assertion to its source. Any claim it cannot ground is stripped, not shipped.
Quality edit
A Pass, Warn, or Block verdict from deterministic checks plus one AI editorial pass. The verdict is computed in code, not by the model, and a Block cannot be approved by a human reviewer.
The quality gate
Grounded, or it does not ship.
Every web-sourced claim's quote must verbatim-match the research it came from. Fabricated or overstated citations fail and block, and a blocked draft cannot be approved. The verdict is computed in code, not by the model, which makes the publish gate predictable and auditable.
- Verbatim claim grounding against captured sources
- Every cited URL present in the Sources block
- Around 50 AI-tell cliche patterns caught and blocked
- Em-dash ban, FAQ, atomic-answer, and table checks
✓claim groundingverbatim match
✓sources coverageall URLs present
✓AI-tell detection~50 patterns
✓em-dash banclean
✓atomic answerspresent
✓comparison tablesper contract
verdict: PASS
# a BLOCK cannot be approved by a reviewer
In the client portal
Versioned brand profile
A guided wizard builds a structured, versioned brand profile (mission, products, ideal customer, differentiators, tone, competitors, target questions). It feeds every stage, and editing it is free.
Monthly content planning
Plan up to 20 posts a month on an interactive board, then turn any planned item into a blog with one click.
Brand and founder voice
Drafts honor your brand and founder voice, banned-word lists, and style constraints, with an editorial voice check on every piece.
Intelligent product placement
Your products are woven in with one of four fit-scored patterns. Featured products are enforced, vetoed products cannot appear.
Two human review gates
You approve the outline and the final draft, with a bounded rewrite allowance. The publish-block is enforced, so a draft that fails grounding cannot be approved.
Export and live status
Export to Markdown, HTML, or Word through short-lived private links. The workspace updates live as a post moves through the pipeline.
End to end
Coming soonFrom approved to published, automatically.
Approve the final draft and Tolstoy takes it the rest of the way. It generates the structured data for the post and publishes it straight to your site, all from inside the tool. No exporting, no copy-paste, no separate SEO pass. Brief to live, end to end.
Automatic schema
Tolstoy generates the JSON-LD for every post (Article, FAQPage, breadcrumbs) so it is machine-readable and eligible for rich results the moment it goes live, no manual markup.
One-click publish
Approve, and the finished post and its schema deploy directly to your connected site or CMS. The work never leaves Tolstoy, and nothing is published until you say so.
Cost control
The system cannot overspend a limit.
Each workspace runs under admin-set limits: a monthly blog quota, a per-blog spend cap, a mandatory monthly ceiling, a rewrite allowance, and a concurrency cap, all enforced before money is spent. Hitting a budget pauses a post at a resumable checkpoint rather than failing it, so no work is lost.
Security
Isolated, invite-only, and audited.
Every customer's data is isolated at the database layer with row-level security. Access is invite-only with strong passwords, mandatory two-factor for operators, multi-layer rate limiting, and a Content-Security-Policy with per-request nonces. Audit logs are append-only.
Tolstoy is rolling out, invite-only.
Want done-for-you content and AI Presence instead? Explore our services.