Prepared for: Sarel
Date: 2026-03-21
What we know so far is pretty clear:
The strongest repeated lesson across the material is this:
- cover the topic well
- answer the actual query fast
- include the important related subtopics and entities
- stop obsessing over exact-match density
Keyword density by itself is not the lever. Good pages naturally include the keyword and relevant variations, but they rank because they are complete, clear, and useful.
Before writing, the page has to match what Google is already rewarding for that query:
- service page
- location page
- comparison page
- FAQ page
- blog post
- listicle
- direct-answer explainer
If the SERP rewards listicles and we write a generic service page, we are fighting the format.
We should keep using light clustering, not giant keyword piles.
That means:
- one main keyword / one main page target
- a few tightly related secondary terms
- supporting questions only when the SERP intent is genuinely shared
This matters even more for local SEO because over-clustering leads to vague pages that do not rank cleanly.
The content has to be easy to parse for:
- human readers
- Google
- featured snippets
- AI Overviews
- ChatGPT / Perplexity / similar tools
That means clean headings, short paragraphs, strong opening answers, lists, tables where helpful, and good internal linking.
The strongest placements are still:
1. Title tag
2. H1
3. First paragraph / first 100 words
4. Keyword variations in H2/H3s
5. Strong topical coverage in body copy
The key point is still: depth should follow intent and coverage, not arbitrary word count goals.
For local businesses, the page has to do two jobs at once:
1. rank for the right local intent
2. convert the visitor into a call, form fill, or estimate request
So local pages cannot just be informational fluff. They need ranking signals and buying signals.
For a local service business, the content system should usually include:
- core service pages
- service + city pages for priority markets
- FAQ pages
- blog/resource pages supporting buyer questions
- local proof pages / project examples where relevant
The best consistent model is:
service × geography × intent
Examples:
- deck builder tampa
- patio contractor brandon
- paver installation near me
- how much does a paver patio cost in tampa
That framework is better than just chasing volume, because local SEO wins by matching real services to real places and real buyer intent.
A strong local service page should usually have:
- benefit-driven headline
- clear value proposition near the top
- strong CTA above the fold
- proof / trust signals
- service explanation
- process or scope details
- testimonials
- FAQs
- final CTA
This is one of the clearest lessons in the whole library:
city pages cannot just swap place names.
A strong local city/service page needs real differentiation, such as:
- local service details
- city-specific testimonials
- city-specific project examples
- local landmarks / neighborhood references
- local FAQ questions
- material or climate notes relevant to that area
- map / NAP / schema where appropriate
A good rule: if you can paste the same content onto another city page and it still reads as true, it is not unique enough.
Mass-produced city pages are risky. What gets us into trouble:
- identical pages with city names swapped
- pages that are not real destinations
- orphaned pages that only exist for keyword capture
- no meaningful local differentiation
Safer approach:
- start with the top real markets
- build fewer, better pages
- make each page a real page with local value
- expand only when the first set works
GBP is important, but it is not enough by itself to rank across the full service area.
That means:
- the verified GBP market is strongest nearby
- surrounding cities still need organic website pages
- reviews, review responses, and website content should reinforce the same service/location themes
Review language is not just social proof anymore. It also helps local relevance and AI visibility.
The content should use:
- real testimonials
- location references
- project outcomes
- review snippets where allowed
- photos / examples / case evidence
The repeated takeaway is:
- Google does not penalize content just because AI helped write it
- Google does punish scaled, low-effort, manipulative content
So the real question is not "AI or no AI?"
The real question is:
Did the page show effort, usefulness, originality, and trust?
Best practice for us:
1. human chooses topic / intent
2. AI helps with research synthesis and drafting
3. AI helps create outline and structure
4. human/editor injects proof, specifics, brand voice, examples, and accuracy
5. final QA checks links, claims, formatting, and intent fit
Common problems:
- generic intros
- hedging language
- repetitive structure
- no real experience
- no original examples
- no local proof
- no source discipline
- no internal link intelligence
We keep seeing the same upgrades matter:
- definitive positions
- specific details
- named places, numbers, tools, products, and outcomes
- original screenshots, photos, or examples
- better paragraph rhythm
- less fluff
- stronger internal linking
- real citations and supporting sources
When using AI for drafts, we should deliberately:
- vary sentence length
- remove cliché AI phrases
- cut filler
- use concrete facts and names
- add firsthand or company-specific details
- include real examples and limits, not just positives
This is a separate but related game.
We are not only trying to rank in classic SERPs anymore. We also want to be:
- cited in AI Overviews
- cited in ChatGPT / Perplexity / other answer engines
- aligned with the format LLMs pull from
One of the best ideas from the AI SEO workflow work is:
- do not only track keywords
- track the actual prompts people ask LLMs
That means content planning can be:
prompt intent → cited-source pattern → preferred content format → article
If AI systems keep citing:
- listicles
- direct-answer explainers
- comparison pages
- FAQ-rich pages
then we should match that format for similar intents.
This is one of the strongest reusable writing patterns.
An answer capsule is:
- a heading that frames a question or subtopic
- followed immediately by a short direct answer paragraph
That helps with:
- featured snippets
- AI Overviews
- voice search
- answer extraction by LLMs
A lot of cited content wins because it answers the core query early.
Best practice:
- short summary near the top
- direct answers under headings
- key facts early, not buried
AI systems seem to prefer pages that are easier to trust.
So instead of dumping sources at the bottom only, our content should include:
- inline support for factual claims
- named sources
- original data when possible
- proof assets when available
Pages with stronger entity coverage and factual density are more likely to be useful to both Google and LLMs.
That means we should naturally include:
- place names
- services
- materials
- brands
- tools
- people
- regulations
- relevant concepts tied to the topic
Freshness is not just a news issue anymore.
The patterns show AI-cited pages tend to be fresher than standard organic winners.
So updates matter for both:
- rankings
- AI citation visibility
The report library repeatedly points to structured content helping interpretation.
Important schema types for this lane:
- Article
- LocalBusiness
- Service
- FAQ
- Breadcrumb
- Person where authorship matters
Internal linking keeps showing up as a real lever.
For our use, the simple rule is:
- every content piece should connect to the right service and conversion pages
- every service page should connect to supporting FAQs/blog content
- every location page should sit inside a sensible local architecture
Good internal links should:
- be descriptive
- vary anchor text naturally
- appear contextually in the body
- help both users and crawlers understand page relationships
For local sites, this also helps reinforce service clusters and city clusters.
The strongest system direction we have so far is:
This is a much better model than just "generate article from keyword and post it."
If we boil everything down, the best current rules are: