How It Works

How Aumata actually works.

A production content engine, not a ChatGPT wrapper. Seven systems working together every day: cluster mapping, strategic planning, multi-pass drafting, a four-layer safety stack, git-based publishing, daily rank plus AEO tracking, and an operator console that shows you all of it.

Planning model

Cluster-first, post March 2026 core update

Drafting model

Opus 4.6, multi-pass per format

Review model

Haiku 4.5, 7-dimension adversarial

Data source

DataForSEO primary, 3-tier fallback

Publishing

Git-based, auto-merge below risk 0.5

Tracking

SEO rank plus LLM Mentions, daily

The Machine

Seven systems. Running every day. All visible to you.

01

Topical Authority Mapping

Before we write a single paragraph, the platform maps every cluster where your domain could credibly win. It scores clusters as owned, adjacent, or gap so investment flows to the shortest path to ranked authority.

DataForSEO ranked_keywords is the primary source, with a 3-tier fallback (competitor intersection, keywords_for_site, Opus site analysis). Never fabricated data. Output: 5 to 12 clusters per client.

02

Cluster-First Strategic Planning

After Google's March 2026 core update, depth beats breadth. For each active cluster, the planner writes the mission, buyer-journey coverage, and a concrete publishing blueprint.

1 pillar (3,000 to 5,000 words) plus 6 to 10 supporting pieces. Every piece ships with a committed Information Gain: the original data, mechanic, or insight it must contribute beyond what the SERP already says.

03

Daily Content Generation

Opus 4.6 drafts in multi-pass rounds, with a dedicated prompt per format: pillar, how-to, comparison matrix, AEO answer, case study, benchmark data, deep dive.

Structured H2s, schema types, word count targets, and required sources are enforced at generation time. Not freestyle prose. Every deliverable follows the spec the cluster plan committed to.

04

Four-Layer Safety Stack

Nothing ships because the model felt confident. Every piece has to clear four independent checks before it publishes.

Layer 1: pre-generation site health gate (accessibility, robots, sitemap, HTTPS, author signals). Layer 2: multi-pass rubric scoring plus research sources requirement. Layer 3: Haiku 4.5 adversarial risk classifier across 7 dimensions. Layer 4: post-publish GSC monitoring.

05

Human-in-the-Loop Publishing

Every draft opens as a GitHub pull request. Auto-merge only runs when confidence clears the threshold and risk score is below 0.5. The strategist can revert any piece in one click.

Three-level kill switch: per-piece (revert PR), per-client (pause the pipeline), global (stop the platform). Every merge decision, revert, and pause is audit-logged.

06

Daily Rank + AEO Tracking

Every day the platform pulls SEO rank deltas for every tracked keyword and LLM Mentions data for share-of-voice across ChatGPT, Claude, and Perplexity.

Daily Slack digest, weekly strategic report. The AEO signal is not a dashboard novelty. It feeds back into which clusters the planner prioritizes next month.

07

Operator Admin Console

Full transparency in a single workspace. Every cluster, every keyword, every piece, every risk score, every merge decision, every operator override is visible and editable.

Unlike agencies that operate as a black box, the strategist can activate clusters, change priority, queue a piece for immediate generation, or override the plan narrative from the same screen they monitor from.

Why it matters

Most AI SEO tools produce volume. Aumata produces ranked, cited authority.

Wrapper tools skip the hard parts. They let Opus or Sonnet free-write, skip site health gates, skip research sourcing, skip structured safety review, and dump content straight to a CMS. The result reads fine, ranks for nothing, and pollutes your domain's topical authority at the exact moment Google is re-weighting for Information Gain.

Aumata was built the other way around. A strategist activates the clusters. The platform commits to an Information Gain per piece. Every draft clears four safety layers and opens as a PR. Every rank delta and LLM mention flows back into next month's plan. You see the machine. You can stop the machine. You can change the plan.

Outcomes

From live client deployments, verified via Ahrefs.

112%

organic traffic growth

B2B IT services client, 90 days

275%

top-3 keyword growth

B2B IT services client, 90 days

415%

traffic value increase

B2B IT services client, 90 days

4,260%

growth from near-zero

managed IT services client

Stats reflect outcomes from a multi-client production deployment. Named case studies available under NDA.

Want to see the machine running on your domain?

Thirty-minute call. We map your clusters live, show you the admin console, and walk through a real plan. No deck, no pitch stack.