SEO Playbook · Process

The SEO process, in the order that works

Eighteen phases, four stages, and one dependency chain. Every phase names its owner, timebox, inputs, outputs, and AmICited surface.

amicited.com/seo-playbook/process
Four-stage dependency chain
Stage AP0–P5 Understand
Stage BP6–P9 Decide
Stage CP10–P14 Build
Stage DP15–P17 Prove & iterate
Compress a phase only when its validated output already exists.
Dependencies, not ceremony

Each phase needs the phase before it

Validate data before researching demand, fix discovery paths before publishing at scale, and freeze the starting point before claiming improvement.

  • Research from verified evidence — confirm permissions, filters, events, and missing feeds first.
  • Design before filling — assign one job to each URL before production creates overlap.
  • Measure before changing — preserve a dated baseline and annotate every intervention.
  • Publish into a healthy path — new content cannot compensate for crawl traps or broken canonicals.
Stage A · Understand

Find out what is true before changing anything

Discovery, data access, technical health, agent readability, and a frozen baseline turn dashboards into evidence.

  • Business context first — define value, markets, constraints, and owners.
  • Instrument before interpreting — test every feed and event.
  • Freeze the starting point — preserve traffic, revenue, rankings, citations, and technical state.
Grouped by what they measure
Organic 2 sources
Unpaid search & AI visibility
Paid search 3 sources
Campaigns bought on queries
Eshops 2 sources
Orders, customers, margin
Every report on the site reads this same grouping — nothing gets blended across channels.
DOMAIN SEMANTIC MAP · 340 PROMPTS CLUSTEREDyourdomain.comg2.comreddit.comcapterra.com
Stages B and C · Decide and build

Turn demand into pages, then ship

Research becomes a prioritized map; the map becomes specifications; the specifications become pages, links, entities, and earned authority.

  • One intent, one accountable URL — resolve overlap before commissioning pages.
  • Reusable controls — post types, evidence standards, review gates, and owners make quality repeatable.
  • Authority follows usefulness — earn citations with assets worth referencing.
Stage D · Prove and iterate

Link releases to outcomes honestly

Conversion tracking, reports, annotations, and refresh rules make the next decision explicit.

  • Trace the chain — visibility, visits, leads, revenue, and retention answer different questions.
  • Annotate interventions — record the change and expected outcome before reading the chart.
  • Refresh by evidence — update, merge, expand, redirect, or retire deliberately.
Outcome ledger · one row per annotation
Rewrote /pricing title tag expected: CTR gap closes Met
Shipped 12 GEO articles expected: Citation rate + Met
Migrated checkout flow expected: Uptime unaffected Inconclusive
Added FAQ schema expected: Rich result gain Missed
Why sequence beats checklists

Why the order matters

Every phase produces evidence or a decision consumed later. A completed activity is worthless when its inputs were never validated.

Research before access

Keyword research before Search Console, analytics, CRM, and market access are verified overweights third-party estimates. The team can chase volume while missing queries that already create qualified demand.

Correct order: verify sources, document gaps, then combine first-party queries with market and prompt evidence.

Production before the map

Separate writers can target the same intent under different titles. The result is cannibalization: several URLs compete for one need and split links and authority.

Correct order: assign intent, post type, URL, and parent-child links before drafting.

Reporting before baseline

A report built after releases cannot recreate the starting state. Seasonality, migrations, campaigns, and tracking changes become convenient explanations.

Correct order: freeze dated segments, annotate interventions, then compare like with like.

The crawl-error multiplier
Publishing onto a site with crawl traps, broken links, incorrect canonicals, or rendering failures compounds the problem. Search engines discover more broken paths faster than useful new content. Visible output rises while indexable output does not.
The rule for moving ahead
Compress a phase only when the team can show the required output, its owner, date, and validation method. “We did this last year” is not an output; a current crawl or approved URL map is.
The flagship map

The complete 18-phase SEO process

Timeboxes are starting ranges for a focused small-to-medium site, not promises. Owner means accountable owner. Inputs and outputs are acceptance criteria.

Stage A — Understand

Remove uncertainty: define value, verify evidence, find technical constraints, test human and agent access, and freeze the starting state.

Typical timeline

What lands by day 30, 60 and 90

These are deliverables, not guaranteed ranking gains, and assume prompt access and available engineering and content capacity.

By day 7
Scope and KPIs are signed. Access is mapped, major data breaks are visible, initial crawl and indexation risks are ranked, and performance and agent-readiness checks are running.
By day 14
Source problems are resolved or documented. The baseline is frozen. Research combines first-party queries with prompts, and competitors and gaps are validated.
By day 30
The priority topical map and URL dispositions are approved. Production and QA operate. Critical fixes and the first cluster can ship with links, measurement, and annotations.
By day 90
Several build-measure-refresh cycles are complete. The team compares like segments, reviews early conversion, ranking, and citation evidence, and reprioritizes without restarting strategy.

A vast marketplace, international migration, or six-week access delay will not fit this pace. The timeline is defensible because those dependencies are exposed rather than hidden in a “strategy month.”

Execution inside AmICited

How AmICited supports each phase

AmICited makes inputs, comparisons, interventions, and outcomes observable; accountable specialists still make and approve the decisions.

P1 · Verify feeds
Data Sources shows what powers every report. Open /data-sources before interpretation.
P2–P4 · Test access
Performance Impact connects page speed to citations; AI Accessibility separates permission, fetchability, extraction, WebMCP, and commerce. Open /performance and /accessibility.
P5 · Freeze visibility
AI Visibility records citations, rank, share of voice, and sentiment. Open /visibility.
P6–P7 · Research the market
Prompt Tracking manages buyer questions; Competitor Analysis discovers rivals from co-citations. Open /prompts and /competitors.
P8–P9 · Map topics and inventory
Semantic Map groups prompts and cited pages by meaning. Combine /semantic-map with /audit/freshness.
P10–P13 · Operate production
AI Content Generation creates controlled drafts and SEO Agents run approved workflows. Open /content and /agents.
P15 · Connect work to money
Revenue Attribution traces trials, orders, MRR, or revenue while preserving caveats. Open /revenue and /reports/cockpit.
P16–P17 · Prove and refresh
Reports Hub, Annotation Outcomes, and Content Freshness create cadence and refresh signals. Open /reports and /audit/freshness.
Credibility by constraint

What we do not do

These tactics create activity reports while transferring risk to the client’s domain.

Volume as a goal
We do not promise page counts detached from distinct demand, review capacity, or crawl health. More URLs can create duplication and maintenance debt.
Bought links and manufactured mentions
We do not buy opaque link packages, use private blog networks, or place undisclosed endorsements. Authority that ends with payment is rented risk.
Schema for invisible facts
We do not mark up invented reviews, ratings, authors, FAQs, or claims. Structured data clarifies visible truth; it does not manufacture credibility.
Vanity movement reported as revenue
We do not convert impressions or mentions into certain commercial impact. Each metric keeps its definition and attribution caveats.
Doorway pages and token swapping
We do not generate near-identical pages by replacing a city, service, or product token. Each URL must provide unique decision value.
Fake freshness
We do not update a date without revalidating substance. Refresh means correcting facts, improving coverage, or changing recommendations when evidence changes.
Measurement defined at the end
We define baselines, events, and expected outcomes before launch, so post-launch charts cannot reward whichever explanation sounds best.
AI output as approval
Generation never bypasses sourcing, expert review, legal checks, brand standards, or technical QA. Automation accelerates a governed workflow.
Human judgment and AI agents

Which phases an agent can actually run

Being honest about this matters more than being optimistic. Some phases are mechanical and benefit enormously from automation; others depend on judgment an agent should not be making alone.

Agents are strong on collection and checking

Pulling query data, crawling for status codes and canonicals, diffing indexation against a sitemap, scoring pages against a checklist, spotting pages missing a required element. These are high-volume, well-defined, and tedious.

This is most of the audit surface in phases 2 through 4, and most of the pre-publish QA gate.

Humans own goals and thresholds

What the business actually sells, what a conversion is worth, which competitor matters, what counts as an acceptable trade-off. Discovery sets the frame every later phase inherits, and a wrong frame is not recoverable by better execution downstream.

An agent can gather the inputs for that conversation. It should not be the one deciding the answer.

Dependencies are what make automation safe

The process is ordered because each phase consumes the previous one's output. An agent that respects the order cannot start producing content before the crawl is clean or the topical map exists.

That ordering is the guardrail. Without it, an agent's speed becomes the problem: it publishes onto a broken foundation faster than anyone can notice.

Thresholds turn observations into decisions

Every phase page states what bad looks like in numbers. An agent can measure against a stated threshold; it cannot reliably infer one.

Where a threshold is genuinely contextual, the page says so, and that is the signal to escalate to a person rather than guess.

The phases are documented in enough detail to be executed rather than interpreted — inputs, outputs, tools, thresholds, and a done-when condition per step. Connect them to your own tooling through the AmICited MCP server, and keep the judgment calls where they belong.

FAQ

SEO process FAQ

Every project needs every decision, but not equal effort. Compress a phase only when its current, validated output already exists.

A first pass commonly spans about 90 days, with useful access and risk findings in week one, baselines and opportunities by day 14, and the first production cycle by day 30.

Drafting can overlap, but publication should wait until critical crawl, indexation, canonical, and rendering risks are understood.

One SEO lead owns sequence and tradeoffs. Analytics, engineering, content, PR, and commercial specialists own their accepted outputs.

A phase advances the project and feeds a later decision. A checklist is a reusable control triggered by an event such as publication or migration.

Start with evidence

Find the gap in your process

Connect the data, capture the baseline, and see where your site and competitors stand before another backlog item becomes a campaign.

Run your free visibility check

18 phases from business context to continuous refresh Six phases establish truth, four choose the work, five build the system, and three prove and improve it. Start with discovery

Put your next SEO decision on verified ground

Free check · 7-day trial · no credit card