Introduction
You want predictable results from AI that improve search visibility and revenue. Prompt engineering SEO gives you a system to do that. It aligns prompts, reusable systems and agent workflows with the way search engines and AI answers select and cite content. In this guide you learn patterns that work, how to build a small content agent, how to shape outputs for schema and snippets, and how to measure real impact with clear KPIs. This matters because generic prompts produce generic pages. Your competitors use AI too. The gap now comes from better instructions, stronger evaluation and fast iteration with data. Use this playbook to win classic results and AI Overviews while you keep quality and safety high.
Foundations of prompt engineering for SEO
Foundations of Prompt Engineering for SEO
Goal Build a repeatable system that turns inputs into search ready outputs with sources, structure and clear evaluation.
Core principles
- Start with the search job to be done. Define the intent, the user mode and the main entity list.
- Require sources. Ask the model to cite or quote from your page or from a vetted knowledge base.
- Constrain format. Ask for tables, headings, and JSON for schema when needed.
- Separate system, instruction and data. Keep a stable system message for brand voice and safety, a task instruction for the job, and the data for context.
- Evaluate every run. Score drafts with a rubric for coverage, accuracy and usefulness.
Starter system message
You are an SEO content editor. Follow brand voice. Favor clear headers and short paragraphs. Cite sources from the provided context only. Refuse unsupported claims.
Task instruction pattern
Task Write a section about [topic] for [audience].
Constraints Max 180 words. Include two H3 ideas. Add one internal link with anchor text. Cite at least one source from context.
Return Markdown only.
Prompt patterns that rank
Goal Use patterns that raise quality and consistency across titles, outlines and copy.
Persona pattern
Act as a senior SEO for a mid market B2B SaaS brand. Audience product managers and content leads. Tone practical.
Few shot pattern
Provide two strong examples before the task. The model follows structure and tone from those examples.
Example A
H2 Benefits of server log analysis
...
Example B
H2 How to monitor crawl budget
...
Now write an H2 about internal linking briefs for topic [X].
Evaluator pattern
Use a second prompt to score the first output. Keep it short and strict.
Score this draft from 1 to 5 for accuracy, completeness, clarity. Flag claims without sources. Return JSON with scores and issues.
Constraint list
Give the model a short list. Use verbs and plain words.
Do
Include entities from the list
Map each H2 to a search intent
Produce one FAQ seed per H2
Do not
Invent stats
Use passive voice
Agentic workflows for content
Goal Move from one prompt to a chain where steps check each other. You get speed and control.
Minimal content agent
- Research step. Summarize the top results and extract entities and questions. Save as a brief.
- Draft step. Convert the brief into a first draft with citations and schema JSON.
- Review step. Run the evaluator prompt and fix issues.
- Link step. Suggest internal links from a URL list with anchor text and target.
You can run this chain by hand inside your AI tool. You can also wire it to a CMS or a knowledge base later.
Example chain prompts
Research
Summarize the top results for [topic]. Return a table with title, main angle, entities, questions, and gaps you can fill.
Draft
Write a 900 word draft for [topic] based on this brief. Add two quotes from the source context with links. Produce Article schema JSON at the end.
Review
Evaluate the draft. List factual risks, missing entities and weak headers. Rewrite weak parts.
Link
From this URL list, find three internal link targets. Return anchor text and link target for each H2.
Keyword research and clustering with LLMs
Goal Turn a large keyword list into clean clusters with intent and entities you can use in briefs.
Clustering prompt
Cluster these 500 keywords by topic and intent. Return CSV with cluster, head term, entity list, search intent, and suggested article angle.
Entity expansion prompt
For cluster [X], list related entities and attributes that must appear in the copy to satisfy the intent.
Cannibalization check prompt
Given these existing URLs and target keywords, flag pages that compete. Suggest a merge or redirect plan with preferred canonical target.
Practical example
A B2B SaaS blog imported 12,000 queries from a keyword tool. The team used the clustering prompt to reduce the list to 320 clusters within one hour. They planned 30 pillars and 90 supports and removed 14 overlapping pages. Organic visits rose twelve percent in eight weeks after they fixed cannibalization and published the first ten supports.
On page optimization prompts
Goal Plan titles, meta descriptions, headings, internal links and FAQ seeds that drive clicks and help AI answers cite you.
Title and meta set
Write 10 title tag and meta description pairs for [topic]. Titles 55 to 60 characters. Front load the keyword. Metas 150 to 160 characters. Avoid brand names.
Heading plan
Propose H2 and H3 for [topic]. Map each H2 to a search intent. List the entity you expect to cover under each H3.
Internal link brief
Given these 50 site URLs, suggest five internal links for this draft. Return source H2, anchor text, target URL.
FAQ seed extraction
Read these top results. Extract five question ideas we can answer with citations on our page.
E E A T, factuality and safety
Goal Keep claims accurate and build trust with readers and with AI systems that pick sources.
Claims and citation prompt
List every claim in this draft that needs a citation. Suggest a reliable source and a short quote for each claim.
Author and review model
Add a short author bio with real expertise. Require a subject matter review step before publish. Show last updated dates. Link to source material and standards. Point to primary sources when you can. Use Google Search Central documentation for reference on helpful content. Validate structured data with the Rich Results Test. Use schema.org types for FAQ, HowTo, Product and Article.
GEO and AI Overviews
AI Search Optimization Prompts
Goal Earn citations in AI answers and show up with a useful summary plus depth that helps the user finish the job.
Dual structure method
- Snippet layer. A short answer that resolves the query in plain words. Add a clear definition or a numbered process.
- Depth layer. A follow up that expands with context, examples and links to tools or sources.
Prompt for dual structure
Write a snippet answer for [query] in 60 words with a clear definition or steps. Then write a depth section with examples, risks and links to primary sources.
Monitoring ideas
Log when your brand appears in AI answers for your tracked topics. Track share of voice across these answers and compare to classic results. Add those metrics to your weekly report.
Multilingual and local SEO with prompts
Multilingual Prompt Engineering
Goal Localize pages for European Portuguese and English with the right entities and tone.
PT and EN workflow
- Translate the draft with locale terms. Keep units and date formats correct.
- Ask for a second pass to match brand voice in the target language.
- Add checks for legal terms and claims.
- Keep hreflang tags and slugs aligned.
Localization prompt
Translate this section to European Portuguese. Keep brand voice. Use local terms and examples. Return a list of terms you changed and why.
Technical SEO with LLMs
Goal Use prompts to speed up logs review, regex creation and schema scaffolds.
Log review prompt
Analyze this server log sample. Find crawl spikes, 404s and slow response clusters. Return a list of fixes.
Regex helper
Write a regex that matches query strings with tracking parameters. Explain how to test it.
Schema scaffold
Produce JSON LD for an FAQ section with three questions and answers based on this copy.
Measurement and PromptOps
Goal Treat prompts and agents like a product. Version them, test them and track outcomes.
KPIs that matter
- AI citation share for tracked topics.
- Retrieval rate from your knowledge base.
- Presence in AI Overviews for target topics.
- Time to first draft.
- Edit effort per draft.
- Leads or assisted revenue influenced by content.
Simple formulas
AI citation share = number of answers that cite your domain / total answers checked
Retrieval rate = answers grounded on your sources / total answers generated
Edit effort = final word count minus words kept from the first draft
Prompt experiment design
- Fix one variable per test. Change the pattern or the constraint, not both.
- Run at least five drafts per variant before you compare.
- Use a blinded reviewer to score quality with the evaluator prompt.
- Publish and track a cohort of pages that use the new template.
Version control and approvals
Keep prompt files in a repo. Add a short readme for each template. Record changes and reasons. Use approvals for risky changes. Train the team on privacy and safe use of source data.
How AISO Hub can help
AISO Hub builds systems that improve how you plan, draft, review and measure SEO content in an AI first world.
AISO Audit We review your content and prompts, map gaps, and set clear KPIs.
AISO Foundation We design your base system messages, prompt templates and governance so the team works in one way.
AISO Optimize We upgrade your keyword clustering, briefs, on page prompts and schema outputs.
AISO Monitor We track AI citations and AI Overview presence and show impact on leads and revenue.
Talk to us and we will set up a short plan you can start this month.
Conclusion
Prompt engineering SEO is not a trick. It is a system that makes AI work for your readers and for your goals. You now have patterns that raise quality, a small agent you can run today, and prompts that build titles, briefs, links and schema with less effort. You also have clear KPIs. Track AI citation share, retrieval rate and AI Overview presence next to your classic metrics. Ship fast and review each step with a simple evaluator. Add sources and keep claims tight. When you make this the way your team works, you write fewer words and get stronger results. If you want a partner to build the system and keep it moving, we can help. See the service section and pick the first step that fits your goals.