Editorial intelligence
HAAM Signal turns evidence into durable search assets
A lightweight internal engine for turning market movement, search demand, product launches, technical traces, and customer questions into verified editorial briefs, useful articles, internal links, and refresh decisions.
The product
A DIY alternative to generic AI SEO publishing
HAAM Signal is designed for judgment, not content volume. It collects signals, ranks the best editorial opportunities, prepares a research packet, writes from evidence, checks claims, and keeps the final publish decision human.
Market signals
Evidence packet
HAAM argument
Verified draft
Human approval
Measured refresh
Inputs
The engine starts with signals, not blank-page prompts
The most useful topics usually come from traces that already exist: what people search, what products ship, what companies hire for, what customers ask, and what the site already knows.
Search evidence
Google Search Console queries, ranking pages, impressions, click-through gaps, and pages that are close to breaking through.
Market evidence
Product Hunt launches, TrustMRR revenue traces, founder posts, job listings, funding notes, and public product changes.
Technical evidence
GitHub activity, changelogs, docs, API releases, performance data, accessibility findings, and integration patterns.
Human evidence
Customer questions, sales calls, support pain, field notes, bookmarks, interviews, and personal observations from the work.
Workflow
The system separates research, writing, verification, and publishing
That separation is the moat. A single prompt can produce text. An editorial engine produces a reusable decision record around the text.
01
Collect
Gather signals before choosing a title
The engine starts with raw evidence. A topic only enters the queue when there is a concrete signal: a query, a launch, a company movement, a pricing change, a technical pattern, or a customer question.
02
Score
Rank ideas by strategic value
Each candidate is scored for HAAM relevance, evidence strength, commercial intent, originality, search demand, and longevity. Low-evidence topics are parked rather than forced into articles.
03
Packet
Create an evidence packet
Before writing, the system assembles sources, competing pages, claims, contradictions, possible internal links, and the one argument the article should make.
04
Draft
Write in the HAAM editorial voice
The draft uses the packet only. It should sound serious, useful, and durable: no fake certainty, no generic filler, no shallow keyword posturing.
05
Verify
Separate writing from verification
A second pass checks unsupported claims, date-sensitive statements, weak links, missing citations, and whether the article has a real reason to exist.
06
Publish
Ship through human approval
The system prepares a page, metadata, internal links, related articles, and a social excerpt, but does not publish automatically. Approval is part of the product.
07
Refresh
Use performance as the feedback loop
Search Console, analytics, inquiries, and backlinks show what should be expanded, merged, rewritten, re-linked, or retired.
Scoring
Every article candidate has to earn its place
The goal is to publish fewer, stronger pieces that compound into HAAM's authority graph. Weak ideas stay in the backlog until more evidence appears.
HAAM relevance
Does this strengthen HAAM's positioning in design, AI, performance, accessibility, trust, or product systems?
Evidence strength
Are there primary sources, data points, observable examples, or first-hand material?
Originality
Can HAAM add a useful argument rather than repeat the existing internet?
Search intent
Would someone actively look for this answer, comparison, method, or market map?
Commercial intent
Could the topic lead to a client, product inquiry, audit, teardown, workshop, or tool?
Longevity
Will the piece still be useful after the immediate news cycle?
Outputs
One signal can become a full editorial asset
The system is useful even before publication because it produces the thinking around the article: sources, claims, positioning, links, and refresh logic.
Guardrails
The engine is deliberately conservative
The strongest version of AI-assisted SEO protects trust. It does not turn HAAM into a content farm, and it does not confuse publishing frequency with authority.
- Never publish without human approval.
- Prefer two to four strong pieces per month over mass content volume.
- Reject topics where the evidence packet is weak.
- Mark claims that need verification instead of hiding uncertainty.
- Use primary sources where possible.
- Keep the article useful even if it never ranks.
- Use internal links to clarify the knowledge graph, not to spray links everywhere.
- Refresh durable pages instead of producing endless near-duplicates.
First implementation
The first useful version can be small
Start with a command-line workflow that creates a research packet and draft brief from a topic, source URLs, and the existing HAAM article graph. Add CMS publishing only after the quality loop is reliable.
pnpm haam:signal -- --topic "AI SEO for design systems" --url https://example.com/source