Practical playbook

SEO & AI Search: An Ongoing Strategy Playbook

Build a living SEO and AI-search strategy from your website, offer and available evidence. The first 30 days establish the foundation; each review creates a capacity-aware next-cycle backlog.

Read the playbook Download the skill

Prefer one portable file? Download the plain Markdown fallback.

Purpose

Plan work the evidence can support.

This playbook helps a business decide which existing pages deserve attention, which information gaps require useful content, what an agent can responsibly do, and how shipped work will be verified and carried into the next cycle.

It fits businesses with a live website, a defined offer, and someone who can verify facts and approve changes. The initial 30-day cycle is a foundation, not the lifespan of the strategy. It is less useful when the offer, audience, or permission path is still undefined.

Core distinction

Delivery completion is not a ranking, traffic, lead, conversion, or citation outcome. Track each on its own timeline.

Required inputs

Start with the business, not a keyword export.

The agent should reuse context already supplied and ask only for material gaps.

01
Website

Primary domain and any priority URLs or sitemap.
02
Business and offer

What the business sells, its differentiators, pricing posture, and claims it can support.
03
Audience and geography

Ideal customer, buying situation, languages, locations, and service limits.
04
Objective

The business action the search program should support and the planning horizon.
05
Current pages

Known commercial, educational, comparison, proof, contact, and conversion pages.
06
Constraints

Capacity, approvals, compliance, technical limits, publishing access, and protected claims.
07
Available evidence

Search exports, analytics, CRM observations, customer questions, research, approved examples, and first-party expertise.

Evidence discipline

Know what the tools can prove.

Tool access changes the research path, not the honesty standard.

Inspect connected tools and supplied exports before recommending research work.
Use Search Console, analytics, DataForSEO, Semrush, Ahrefs, or another source only when it is actually connected or the user supplied an export.
Record each source as available, supplied export, unavailable, or not needed for this scope.
Do not treat a general language-model answer as live search-result, indexing, traffic, conversion, or citation evidence.
Reuse supplied context and ask only for material missing information.
Keep observed facts, source estimates, business hypotheses, and recommendations visibly separate.
Label unavailable measurements as Not available, not zero.
Keep search-volume estimates, organic difficulty, ad competition, indexing, traffic, conversions, and sampled AI citations in separate fields.
Keep delivery completion separate from rankings, traffic, leads, citations, and revenue outcomes.
Preserve the receiving user's permissions. Planning or review does not authorize publishing, spending, sending, outreach, or account changes.
Ongoing means the method can be repeated when a user starts a new authorized cycle. It does not authorize scheduled tasks, continuous monitoring, paid tool calls, publishing, or outreach.

Initial 30-day cycle

A foundation cycle that adapts to capacity.

Use this sequence to establish the first implementation baseline. Reduce the number of pages and actions when capacity is tight. Do not remove evidence, ownership, dependencies, or acceptance checks to make the plan look faster.

01 Days 1 to 5

Baseline, research, and technical access

Establish what can be observed, what the site already owns, and which gaps are real before expanding scope.

Inventory current indexable and priority URLs, then assign each a visible audience, intent, offer, and next-step job.
Register available tools and exports. Record collection dates, filters, geography, device, and denominators where relevant.
Build a query-to-existing-URL intent map from customer language, supplied evidence, and live search research when available.
Check technical access, crawl controls, canonicals, status responses, sitemap inclusion, internal discoverability, and important rendered text.
Create a dated baseline without filling unavailable fields with invented values.

02 Days 6 to 12

Improve the right pages and factual entity or offer information

Make the pages that already own buyer needs more accurate, useful, and explicit before creating new inventory.

Prioritize existing pages by business value, evidence gap, technical risk, effort, and confidence in the diagnosis.
Prepare page-improvement briefs covering intent, audience, direct answer, proof, factual entity or offer details, structure, internal links, and next step.
Resolve inconsistent business names, categories, service areas, offer details, contact paths, and supported structured data where the evidence allows.
Keep recommendations scoped to available capacity and approval paths.

03 Days 13 to 22

Create useful sourced answer-ready content and contextual links

Fill verified information gaps with original, useful material and connect it to the pages that need support.

Create or improve only the content required by the intent map and page briefs.
Lead with a direct answer where it helps the reader, then add context, original expertise, sources, limits, and a useful next step.
Add crawlable contextual links with descriptive anchors where the destination genuinely helps the reader.
Attribute borrowed methods and preserve readable original-source URLs.
Do not manufacture mentions, citations, reviews, examples, or scaled pages to imitate authority.

04 Days 23 to 30

Verify shipped changes and review comparable observations

Confirm that authorized work actually shipped, then compare observations without turning correlation into a claimed outcome.

Verify each authorized release on its canonical production URL, including response, crawlability, metadata, structured data, links, accessibility, and responsive rendering.
Record delivery evidence for every planned item and list anything not shipped with its dependency.
Repeat only comparable search or AI-answer samples with the same question, geography, language, device, source, and collection method where possible.
Review indexing, organic discovery, sampled AI citations, referral activity, and conversions as separate observations.
Choose the next cycle from the strongest verified gap, not from a promised ranking or an isolated favorable answer.

Repeat beyond day 30

Turn each review into the next useful cycle.

Update the evidence baseline, review delivery and results separately, carry unfinished work forward, add only newly justified opportunities, and prioritize what fits the available capacity.

01
Update the evidence baseline
Review available first-party data, supplied exports, live technical checks, and dated search observations. Keep unavailable measures as Not available.
02
Review delivery and results separately
Confirm what was delivered and verified before reviewing rankings, traffic, citations, inquiries, or revenue observations on their own timelines.
03
Carry unfinished work forward
Keep each existing work item’s original ID, current status, recorded status history or evidence-transition reference, dependencies, owner, and acceptance checks visible. Never renumber it, overwrite past delivery or status evidence, or auto-complete it because a target date passed. Give proposed new work distinct proposed IDs.
04
Add newly justified opportunities
Add an opportunity to the maintained backlog only when current evidence and available capacity make it specific enough to scope.
05
Prioritize the next capacity-aware cycle
Choose the smallest useful set of next actions, with owners, dependencies, effort, target URLs, and delivery acceptance evidence.
06
Append a dated decision and change log
Record the evidence reviewed, delivery status, result status, lessons, carry-forward work, and the next review checkpoint.

Permission boundary

Ongoing does not authorize scheduled tasks, continuous monitoring, paid tool calls, publishing, or outreach. Each action still requires the permissions of the person or system using the playbook.

What you will create

Require work another person can review.

Concise business brief

business and offer
audience and geography
objective
priority conversion action
constraints
known evidence
unresolved material inputs

Tool and data-availability register

source
status
scope
date range
filters or denominator
what it can support
what remains unavailable

Query-to-existing-URL intent map

query or question family
intent
audience or location
existing primary URL
page job
evidence
gap
recommended action

Prioritized foundation or current/next capacity-aware cycle

existing work-item ID or proposed new ID
current status
recorded status history or evidence-transition reference
window or cycle
action
owner
reviewer
dependency
effort
target URL
delivery acceptance evidence
outcome observation and timing

Page-improvement briefs

target URL
current job
observed issue
recommended change
facts or sources required
contextual links
approval need
acceptance check

Measurement baseline

measurement lane
value or Not available
source
collection date
scope
denominator or filters
limitation
next comparable review

Review log

date
decision
owner
evidence reviewed
delivery review
results review
lesson
next action

Next-cycle backlog and review checkpoint

existing work-item ID or proposed new ID
current status
recorded status history or evidence-transition reference
carried-forward task
newly justified opportunity
scope evidence
priority
capacity
owner
dependency
acceptance check
review date or trigger

Measurement

Seven lanes. No blended visibility score.

A value must keep its source, collection date, scope, filters or denominator, and limitation.

Search-volume estimate

An estimate from a named source and scope. It is not observed site traffic.

Organic difficulty

A vendor-specific estimate. It is not Google data or a guarantee of ranking effort.

Ad competition

Paid-search advertiser competition. It is not organic ranking difficulty.

Indexing

Observed or reported eligibility and index state. It is not ranking or traffic.

Organic traffic

Visits or clicks attributed by the named source and filters. It is not a conversion.

Conversions

Defined business actions with a named tracking source. They do not prove SEO caused the result.

Sampled AI citations

Dated observations from a fixed question and environment. They are directional, not a universal rank.

Working example

See how Markethink logs its own implementation.

Markethink's living public SEO and AI-search strategy shows decisions, sources, shipped work, lessons, and review notes on a live site. It is Markethink's implementation record. Use it to understand the operating rhythm, not as independent proof and never as a source of statuses for another business.

Open the public implementation log

Use with an agent

Give the method enough context to stay useful.

Choose the format your agent supports. Copy and paste works broadly. Markdown works as a self-contained attachment. Native skill import is only for products that support the public specification.

Copy and paste

Paste the complete standalone brief into any capable agent and keep your business context with it.

Attach Markdown

Download one self-contained Markdown file when an agent handles attachments or long context more reliably.

Import the skill

Use the ZIP only where the agent or product supports the Agent Skills specification and skill import.

View or manually copy the complete standalone brief

Complete standalone brief

SEO & AI SEARCH: ONGOING STRATEGY BRIEF
Version: 1.1.2
Reviewed: 2026-09-13
Source: https://markethink.ai/ai-marketing-strategies/seo-aio-playbook/

ROLE
Help me adapt and maintain an evidence-led SEO and AI-search strategy for my business. For a new program, produce the initial 30-day foundation cycle. For an established program, resume from its existing records and produce the current or next capacity-aware cycle without restarting day 1. Support planning, review, and only the execution I separately authorize. Ongoing does not authorize scheduled tasks, continuous monitoring, paid tool calls, publishing, or outreach. It also does not authorize spending, sending, account changes, or any action outside my permissions.

OPERATING RULES
1. Reuse supplied context and ask only for material missing information.
2. Keep observed facts, source estimates, business hypotheses, and recommendations visibly separate.
3. Label unavailable measurements as Not available, not zero.
4. Keep search-volume estimates, organic difficulty, ad competition, indexing, traffic, conversions, and sampled AI citations in separate fields.
5. Keep delivery completion separate from rankings, traffic, leads, citations, and revenue outcomes.
6. Preserve the receiving user's permissions. Planning or review does not authorize publishing, spending, sending, outreach, or account changes.
7. Ongoing means the method can be repeated when a user starts a new authorized cycle. It does not authorize scheduled tasks, continuous monitoring, paid tool calls, publishing, or outreach.

CONTEXT TO COLLECT
Reuse anything I already supplied. Ask only for material missing information, grouped into one concise request when possible.
- Website: Primary domain and any priority URLs or sitemap.
- Business and offer: What the business sells, its differentiators, pricing posture, and claims it can support.
- Audience and geography: Ideal customer, buying situation, languages, locations, and service limits.
- Objective: The business action the search program should support and the planning horizon.
- Current pages: Known commercial, educational, comparison, proof, contact, and conversion pages.
- Constraints: Capacity, approvals, compliance, technical limits, publishing access, and protected claims.
- Available evidence: Search exports, analytics, CRM observations, customer questions, research, approved examples, and first-party expertise.

TOOL AND DATA CHECK
- Inspect connected tools and supplied exports before recommending research work.
- Use Search Console, analytics, DataForSEO, Semrush, Ahrefs, or another source only when it is actually connected or the user supplied an export.
- Record each source as available, supplied export, unavailable, or not needed for this scope.
- Do not treat a general language-model answer as live search-result, indexing, traffic, conversion, or citation evidence.

FOUNDATION OR CURRENT/NEXT IMPLEMENTATION CYCLE
For a new program, use the initial 30-day sequence as the foundation cycle, not the lifespan or completion deadline of the strategy. For an established program, preserve its records and use these phases only as a method for the current or next capacity-aware cycle; do not restart day 1.
Days 1 to 5: Baseline, research, and technical access
  - Inventory current indexable and priority URLs, then assign each a visible audience, intent, offer, and next-step job.
  - Register available tools and exports. Record collection dates, filters, geography, device, and denominators where relevant.
  - Build a query-to-existing-URL intent map from customer language, supplied evidence, and live search research when available.
  - Check technical access, crawl controls, canonicals, status responses, sitemap inclusion, internal discoverability, and important rendered text.
  - Create a dated baseline without filling unavailable fields with invented values.
Days 6 to 12: Improve the right pages and factual entity or offer information
  - Prioritize existing pages by business value, evidence gap, technical risk, effort, and confidence in the diagnosis.
  - Prepare page-improvement briefs covering intent, audience, direct answer, proof, factual entity or offer details, structure, internal links, and next step.
  - Resolve inconsistent business names, categories, service areas, offer details, contact paths, and supported structured data where the evidence allows.
  - Keep recommendations scoped to available capacity and approval paths.
Days 13 to 22: Create useful sourced answer-ready content and contextual links
  - Create or improve only the content required by the intent map and page briefs.
  - Lead with a direct answer where it helps the reader, then add context, original expertise, sources, limits, and a useful next step.
  - Add crawlable contextual links with descriptive anchors where the destination genuinely helps the reader.
  - Attribute borrowed methods and preserve readable original-source URLs.
  - Do not manufacture mentions, citations, reviews, examples, or scaled pages to imitate authority.
Days 23 to 30: Verify shipped changes and review comparable observations
  - Verify each authorized release on its canonical production URL, including response, crawlability, metadata, structured data, links, accessibility, and responsive rendering.
  - Record delivery evidence for every planned item and list anything not shipped with its dependency.
  - Repeat only comparable search or AI-answer samples with the same question, geography, language, device, source, and collection method where possible.
  - Review indexing, organic discovery, sampled AI citations, referral activity, and conversions as separate observations.
  - Choose the next cycle from the strongest verified gap, not from a promised ranking or an isolated favorable answer.

REPEAT CYCLE BEYOND DAY 30
1. Update the evidence baseline: Review available first-party data, supplied exports, live technical checks, and dated search observations. Keep unavailable measures as Not available.
2. Review delivery and results separately: Confirm what was delivered and verified before reviewing rankings, traffic, citations, inquiries, or revenue observations on their own timelines.
3. Carry unfinished work forward: Keep each existing work item’s original ID, current status, recorded status history or evidence-transition reference, dependencies, owner, and acceptance checks visible. Never renumber it, overwrite past delivery or status evidence, or auto-complete it because a target date passed. Give proposed new work distinct proposed IDs.
4. Add newly justified opportunities: Add an opportunity to the maintained backlog only when current evidence and available capacity make it specific enough to scope.
5. Prioritize the next capacity-aware cycle: Choose the smallest useful set of next actions, with owners, dependencies, effort, target URLs, and delivery acceptance evidence.
6. Append a dated decision and change log: Record the evidence reviewed, delivery status, result status, lessons, carry-forward work, and the next review checkpoint.

REQUIRED OUTPUTS
Return every section below. Use tables where they improve readability.
1. Concise business brief: business and offer; audience and geography; objective; priority conversion action; constraints; known evidence; unresolved material inputs
2. Tool and data-availability register: source; status; scope; date range; filters or denominator; what it can support; what remains unavailable
3. Query-to-existing-URL intent map: query or question family; intent; audience or location; existing primary URL; page job; evidence; gap; recommended action
4. Prioritized foundation or current/next capacity-aware cycle: existing work-item ID or proposed new ID; current status; recorded status history or evidence-transition reference; window or cycle; action; owner; reviewer; dependency; effort; target URL; delivery acceptance evidence; outcome observation and timing
5. Page-improvement briefs: target URL; current job; observed issue; recommended change; facts or sources required; contextual links; approval need; acceptance check
6. Measurement baseline: measurement lane; value or Not available; source; collection date; scope; denominator or filters; limitation; next comparable review
7. Review log: date; decision; owner; evidence reviewed; delivery review; results review; lesson; next action
8. Next-cycle backlog and review checkpoint: existing work-item ID or proposed new ID; current status; recorded status history or evidence-transition reference; carried-forward task; newly justified opportunity; scope evidence; priority; capacity; owner; dependency; acceptance check; review date or trigger

MEASUREMENT SEPARATION
Never collapse these into one visibility score.
- Search-volume estimate: An estimate from a named source and scope. It is not observed site traffic.
- Organic difficulty: A vendor-specific estimate. It is not Google data or a guarantee of ranking effort.
- Ad competition: Paid-search advertiser competition. It is not organic ranking difficulty.
- Indexing: Observed or reported eligibility and index state. It is not ranking or traffic.
- Organic traffic: Visits or clicks attributed by the named source and filters. It is not a conversion.
- Conversions: Defined business actions with a named tracking source. They do not prove SEO caused the result.
- Sampled AI citations: Dated observations from a fixed question and environment. They are directional, not a universal rank.

ACCEPTANCE RULES
- Adapt the number of pages, queries, and actions to the stated capacity.
- Every action must have an owner, reviewer where needed, dependency, effort, target URL or asset, and delivery acceptance evidence.
- Mark unavailable data as Not available and state what would be needed to measure it.
- Treat a completed deliverable as delivery evidence only. Keep rankings, traffic, leads, conversions, and citations pending until observed from a named source.
- Review delivered work and results in separate fields.
- Carry unfinished work forward with its original work-item ID, current status, recorded status history or evidence-transition reference, dependency, and acceptance check. Never renumber it, overwrite past delivery or status evidence, or auto-complete work because a date passed.
- Label newly justified work with distinct proposed new IDs so it cannot be mistaken for or overwrite an existing record.
- Add a newly justified opportunity only after its evidence, scope, priority, capacity, owner, dependency, and acceptance check are explicit.
- End every cycle with a next-cycle backlog, a review date or trigger, and an appended dated decision/change log.
- Preserve original source URLs and attribute any borrowed strategy.
- Do not present Markethink's roadmap or any illustrative example as proof for this business.
- Do not schedule the next review, create a monitor, use a paid tool, publish, or begin outreach unless I separately authorize that action.

LIMITATIONS
- Search and AI-answer systems change. A plan can improve clarity, crawlability, usefulness, and evidence without guaranteeing selection, ranking, citation, traffic, or leads.
- Low-volume businesses may need review windows longer than 30 days for comparable outcome observations.
- Tool estimates use different datasets and definitions. Record the provider and do not merge unlike metrics.
- The first 30 days are an initial implementation cycle, not the lifespan of the strategy or a claim that search engines will recrawl, index, rank, or cite a page on that schedule.
- A current baseline is evidence for prioritization, not a permanent cap on future actions. New work enters the maintained backlog only when it is justified and scoped.
- Markethink's public roadmap is an implementation example from Markethink.ai. It is not independent proof and its statuses must not be copied into another business's plan.

ORIGINAL SOURCES
- Google Search Central, Optimizing your website for generative AI features on Google Search: https://developers.google.com/search/docs/fundamentals/ai-optimization-guide
- Google Search Central, AI features and your website: https://developers.google.com/search/docs/appearance/ai-features
- Google Search Central, Creating helpful, reliable, people-first content: https://developers.google.com/search/docs/fundamentals/creating-helpful-content
- Google Search Central, Google Search's guidance on using generative AI content on your website: https://developers.google.com/search/docs/fundamentals/using-gen-ai-content
- Google Search Central, Link best practices for Google: https://developers.google.com/search/docs/crawling-indexing/links-crawlable

START
First summarize the context already available. Then show the tool and data-availability register. Ask only for material missing inputs. If this is a new program, produce the initial foundation cycle. If this is an established program, resume from existing records and produce the current or next capacity-aware cycle without restarting day 1. Review delivery and results separately, preserve original work-item IDs and status history, distinguish proposed new IDs, and finish with an honest next-cycle backlog and review checkpoint.

Download Markdown

Limitations

Useful work without outcome promises.

01
Search and AI-answer systems change. A plan can improve clarity, crawlability, usefulness, and evidence without guaranteeing selection, ranking, citation, traffic, or leads.
02
Low-volume businesses may need review windows longer than 30 days for comparable outcome observations.
03
Tool estimates use different datasets and definitions. Record the provider and do not merge unlike metrics.
04
The first 30 days are an initial implementation cycle, not the lifespan of the strategy or a claim that search engines will recrawl, index, rank, or cite a page on that schedule.
05
A current baseline is evidence for prioritization, not a permanent cap on future actions. New work enters the maintained backlog only when it is justified and scoped.
06
Markethink's public roadmap is an implementation example from Markethink.ai. It is not independent proof and its statuses must not be copied into another business's plan.

Original sources

Read the guidance behind the method.

Reviewed 2026-09-13. Source pages can change, so read the live guidance before relying on a detail.

Google Search Central Optimizing your website for generative AI features on Google Search
Foundational SEO, valuable non-commodity content, technical clarity, supported measurement, and limits on AI-search shortcuts.
https://developers.google.com/search/docs/fundamentals/ai-optimization-guide
Google Search Central AI features and your website
Eligibility, crawlability, indexability, internal links, visible text, structured-data consistency, and reporting limits.
https://developers.google.com/search/docs/appearance/ai-features
Google Search Central Creating helpful, reliable, people-first content
Originality, completeness, sourcing, authorship, audience value, and avoiding search-engine-first production.
https://developers.google.com/search/docs/fundamentals/creating-helpful-content
Google Search Central Google Search's guidance on using generative AI content on your website
Accuracy, quality, relevance, transparency, and scaled-content limits when AI supports production.
https://developers.google.com/search/docs/fundamentals/using-gen-ai-content
Google Search Central Link best practices for Google
Crawlable anchors, descriptive link text, contextual internal links, and useful external citations.
https://developers.google.com/search/docs/crawling-indexing/links-crawlable