20 Best ChatGPT Prompts for SEO - November 2026

The secret to good SEO output from ChatGPT is feeding it data and saying clearly what it must not do. The 20 prompts below come from patterns I use again and again in my own projects. Fill the fields in square brackets with your own data.
I explain where ChatGPT helps and where it misleads in SEO with ChatGPT. This post focuses on ready-to-use prompts, and under each one I note what to verify the output against.
Short answer: The best SEO prompts include context, real data, a clear task, an output format and constraints. Use ChatGPT for keyword classification, content briefs, meta suggestions, JSON-LD drafts, regex, internal links and competitor comparison, not for volume, rankings or statistics. Verify every output with an official tool or your own data.
How do you build a good SEO prompt?
Every prompt in this list uses the same five parts. You can use this structure as a template for your own prompts.
- Suggest keywords for SEO
- No data, no context
- Unclear format
- The tool may invent volume
- Classify the attached 200 queries by intent
- Search Console data attached
- Table format requested
- Includes a mark-uncertain-rows constraint
Keyword research prompts
1. Classify by search intent
You are an experienced SEO specialist. Classify the keyword list below by search intent: informational, commercial investigation, transactional, navigational. Return a table: keyword | intent | one-sentence reason. Mark rows you are unsure about as 'unclear'.
List:
[PASTE KEYWORDS]Verify: Search the 'unclear' queries on Google and look at the result types on page one. What Google shows for a query is what really defines its intent.
2. Group keywords for one page
Group the keywords below into sets that can be targeted by a single page. Merge queries with the same intent that one piece of content can answer. For each group, give a primary keyword and a suggested page type (blog, category, product, service). Do not estimate search volume.
[LIST]Verify: If two queries share most of their page-one results, one page makes sense. The full method is in the keyword research guide.
3. Generate long-tail questions
Generate 30 questions that users in [COUNTRY] might type into Google about [TOPIC]. Use everyday language. Show common misspellings and variants in a separate column. Split the questions into informational and purchase intent.Verify: The questions are an idea list. Check whether people actually search them with Search Console, Google Trends or a keyword tool.
4. Search Console opportunity queries
The attached Search Console query export has these columns: query, clicks, impressions, CTR, average position. Find 20 queries with an average position between 4 and 15, high impressions and a CTR below other queries at similar positions. For each, give a likely cause (title mismatch, intent gap, SERP feature) and a suggested action.Verify: Filter the chosen queries in Search Console and check the numbers one by one. The Search Console guide explains how to read the reports.
Content draft and brief prompts
5. Content brief
Prepare a content brief for [KEYWORD]: target reader, search intent, suggested H2 and H3 structure, questions to answer, and ideas for original examples or data to add. Do not invent any numbers or statistics; mark places that need a source as [SOURCE NEEDED].Verify: Compare the brief with page-one results and decide what experience you can add. The content strategy guide clarifies which topic cluster the brief belongs to.
6. Short answer paragraph
The main question of the article below is: [QUESTION]. Write a 2-3 sentence 'short answer' paragraph that answers it directly with a definition and the main recommendation. No marketing language or hype.
[TEXT]Verify: The paragraph must not contradict the rest of the text. Every claim in the short answer should be supported in the body.
7. Find gaps in old content
Review the article and target query below. List missing subtopics, statements that may be outdated (dates, versions, feature names) and repeated sections as bullet points. Do not rewrite the text, only give the list.
Query: [QUERY]
Text: [TEXT]Verify: Check every statement flagged as outdated against an official source; the tool can mistake current facts for old ones.
8. FAQ questions
Suggest 6 questions a reader might genuinely ask that the body of the article below does not answer. Write a clear 2-3 sentence answer for each. Do not repeat questions the body already answers.
[TEXT]Verify: Correct the answers with your own expertise. FAQPage markup remains valid, but since May 2026 Google no longer shows FAQ rich results in Search.
Meta title and description prompts
9. Title tag variations
Suggest 5 different title tags for the page below. Keep the primary keyword near the start and the brand name at the end, and keep it short enough not to be truncated in search results. Give each with its character count.
Page topic: [TOPIC]
Primary keyword: [KEYWORD]
Brand: [BRAND]Verify: ChatGPT can miscount characters. Test the options in the SERP preview tool; Google may still rewrite the title.
10. Meta description
Write 3 meta descriptions for the page below that encourage clicks and summarize information that is actually on the page. Keep each under 150 characters with no exaggerated promises.
[PAGE TEXT]Verify: Every promise in the description must be backed up on the page. The full rules are in meta title and description rules.
11. Bulk category meta tags
The attached CSV has URL, category name and parent category columns. Generate a title and a meta description for each row. Do not repeat the same pattern; use a detail specific to each category. Return a CSV with the same columns plus two new ones.Verify: Read the first 20 rows by hand; repeated patterns and wrong category details are the most common problems.
Schema structured data prompts
12. JSON-LD draft
Generate LocalBusiness JSON-LD with the details below. Use only the fields I provide; do not add or guess any field you do not know.
Name: [NAME]
Address: [ADDRESS]
Phone: [PHONE]
Opening hours: [HOURS]
URL: [URL]Verify: Check the code in Google's Rich Results Test. For a ready-made alternative, use the schema generator.
13. Debug existing markup
Review the JSON-LD below for syntax errors, wrong property names and values that do not match the visible content of the page. List the errors, then write the corrected code.
[CODE]Verify: Confirm which properties each type needs in Google's structured data documentation; the schema and structured data guide walks through this check.
Regex and data analysis prompts
14. Regex for question queries
Write a regex for the Google Search Console performance report filter that matches question queries. Cover: how, what, why, when, where, which, who, can, does, is. Explain what it matches and what it might miss.Verify: Test the regex on a small query list first. Short words such as 'is' can match queries you do not want; check whether word boundaries are used.
15. Separate branded and non-branded queries
My brand name is [BRAND]. Write a regex for the Search Console regex filter that isolates branded queries. Cover common misspellings, spaced and joined spellings, and versions with and without accents.Verify: After applying the filter, read the top 50 non-branded queries; add any missed spelling to the regex.
16. Google Sheets formula
In Google Sheets, column A has full URLs. Write a REGEXEXTRACT formula for column B that extracts the first folder of the URL (for example /blog/). Leave it blank if there is no folder. Explain how the formula works.Verify: Test the formula on a root URL, a URL with parameters and a URL without a trailing slash.
Internal linking prompts
17. Internal links for a new article
Below is a list of URLs and titles on my site, followed by the text of my new article. Pick the 5 most relevant pages to link to from the new article. For each, suggest the right sentence and a natural, keyword-rich anchor text. Do not use phrases like 'click here'.
Pages: [LIST]
Article: [TEXT]Verify: Check that the suggested URLs really exist in your list; the tool sometimes invents addresses. See internal linking for the principles.
18. Pages with few links
The attached crawl export has URL, title and inlinks columns. List pages with fewer than 3 internal inlinks and suggest the 3 closest source pages by topic for each.Verify: Match the suggestions against pages that bring traffic; not every page with few links is a priority.
Competitor analysis prompts
19. Compare competitor heading structure
Below are the H2 and H3 headings of three competitor pages ranking on page one for the same query. List in three groups: topics all of them cover, topics only one covers, and topics none cover that users might ask about.
[HEADINGS]Verify: Copy the headings from the pages yourself; asking the tool to 'read' a competitor page can return incomplete or outdated results.
20. Compare service page value propositions
Below are my service page copy and two competitors' pages. Compare value proposition, trust elements (testimonials, certifications, process description), how pricing information is presented, and calls to action in a table. List what my page is missing separately.Verify: The table is a starting point; base decisions on client conversations and conversion data. The full method is in SEO competitor analysis.
What should you verify prompt output with?
| Category | Data to feed the prompt | Verification tool |
|---|---|---|
| Keywords | Search Console queries, tool exports | Google results page, Search Console |
| Content drafts | Existing copy, competitor headings | Expert review, official sources |
| Meta tags | Page copy, category list | SERP preview |
| Schema | Business details, existing code | Rich Results Test |
| Regex and formulas | Sample query and URL lists | Test on a small sample |
| Internal links | URL and title list, crawl data | Crawler |
- Does every number in the output come from your own data?
- Do the suggested URLs actually exist on your site?
- Have code and regex been tested on a small sample?
- Does the text include your own experience and examples?
I summarize what Google expects when you publish AI output in AI content and Google's policy. For options beyond ChatGPT, see the list of AI-powered SEO tools.
If you want to fit these prompts into your content process or build a workflow for your team, see content strategy services or book an intro call.
Frequently Asked Questions
Do these prompts work in assistants other than ChatGPT?
Yes, the structure is the same, so you can use them in Gemini, Claude or Copilot. Output format and length may vary by tool. Running the same prompt in two tools and comparing results is a good quality check.
Should I write prompts in the language of my content?
For content tasks, prompting in the target language usually gives more natural results. For technical tasks such as regex, JSON-LD or formulas, language does not matter. State the output language explicitly in the prompt.
Can I trust search volume figures from ChatGPT?
No. General assistants do not provide reliable volume data on their own and can invent numbers. Take volume from Google Ads Keyword Planner or a keyword tool, and use ChatGPT to interpret it.
Why does the same prompt give different results each time?
Language models produce slightly different answers on each run. For consistency, define the format and constraints tightly and include a sample output row. Test bulk jobs on a small sample first.
How should I share prompts with my team?
Keep your prompts in a shared document with category, purpose and verification step. Note what changed in the output after each update. That way the team gets the same quality repeatably.

8+ years of hands-on SEO and GEO experience across 150+ projects in e-commerce, SaaS, healthcare and local business sectors. Writes based on real client data, not theory.





