30 AI Prompts for SEO: Templates to Copy, What Each Does and Its Limit

30 AI prompts for SEO in copy-ready blocks, from keywords to reports, with what each one does, its limit, what not to paste and what Google says about AI content.
Written with AI, reviewed by humans

In this article

See how your brand shows up in AI answers

Book a demo and find out what to fix first, with a GEO specialist.

Book a demo

Reading time: 11 minutes

AI prompts for SEO are ready-made instructions that you paste into an AI like ChatGPT, Claude or Gemini to speed up a search task. Below are 30 AI prompts for SEO, from keywords to reports, with what each does and its limit. AI speeds up the draft, but you check the facts and numbers.

In short

  • The 30 AI prompts for SEO cover research, briefs, titles, audits, FAQs, schema and reports.
  • AI can get facts and math wrong. So check numbers, dates and names before you use the answer.
  • As good practice, do not paste client data, because training policy changes by tool and plan.
  • Google sees value in AI but asks for review, and mass pages with no value can break its spam policy.

How do you use these AI prompts for SEO without mistakes?

First, replace the parts in brackets with your data, run one prompt at a time and check the answer before you use it. These AI prompts for SEO are our own templates, for any AI chat. We did not test them in every tool, and the limits below are our own cautions. The answer changes from one tool, or one run, to another, so treat it as a draft. For the big picture, see the guide to AI for SEO and GEO.

So every prompt follows the same template, in four lines:

Context: [company, audience and location, such as a Denver HVAC company serving local homeowners]
Task: [what the AI should do]
Format: [table, list, 5 items]
Rules: use only the data I pasted. If information is missing, say what is missing. Do not invent numbers, sources or links.

For that reason, use the rules line in every prompt: the AI says so when data is missing.

AI prompts for SEO: illustration of a panel with ready-to-copy AI prompt blocks

What should you not paste into AI prompts for SEO?

As good practice, do not paste client data into AI prompts for SEO: names, emails, contracts, passwords, API tokens and exports with personal data. Instead, paste only public text and aggregated numbers, with the client’s permission, because training policy changes by tool and plan.

The official pages cover training and retention:

  • Claude (Free, Pro and Max): according to Anthropic, chats can train the model if you allow it, and chats flagged for safety review can also be used. However, Incognito chats are not used. In Claude for Work and the API, the default is no training, per the commercial policy.
  • ChatGPT (Free, Plus and Pro): according to the OpenAI help page, your content can train the models unless you turn off “Improve the model for everyone”. On the other hand, Business, Enterprise, Edu and the API do not train by default.
  • Gemini: the Privacy Hub says human reviewers read some of the data and reviewed chats are kept for up to three years. So it asks you to avoid confidential information.

Even with training off, anonymize the client, since your company’s contract may require more.

Which AI prompts suit keyword research?

The ones that organize and expand ideas suit it, not the ones that measure search. So the AI lists likely searches, groups them by intent and separates questions. However, volume comes from Search Console or a keyword tool. Without that data pasted in, the AI only estimates (our own assessment, no source).

1. Search ideas by stage. What it does: lists the searches your audience makes before buying. Limit: they are hypotheses, because the AI does not measure volume.

I am a [job title] at a company that sells [product or service] to [audience, such as independent dental practices in Austin, Texas].
List 20 searches this audience would type into Google before buying, in American English, split into three stages: discovery, comparison and decision.
Answer in a table with two columns: search and stage.

2. Group by intent. What it does: splits the list into information, comparison, purchase and brand. Limit: the AI sorts by the text, but only the results page shows what Google returns.

Group the keyword list below by search intent: information, comparison, purchase and brand.
Show a table: keyword, intent and one sentence explaining the choice.
List separately the ones you were unsure about.

[paste the list]

3. Audience questions. What it does: lists questions that become section headings. Limit: they are plausible, not measured, so compare them with Search Console.

Topic: [article topic]. Audience: [audience, such as small-business owners in the US].
List 12 questions this audience would ask about the topic, from the most basic to the most advanced.
Mark which ones work as a section heading (H2) in an article.

4. Variations and ways to ask. What it does: brings synonyms, acronyms and other ways to say the same thing. Limit: the AI can invent slang, so test the terms in Google search.

For the search "[keyword]", list synonyms, acronyms, plurals and different ways to ask the same thing in American English.
Split into two lists: terms I can use in the text and terms that look like another topic.

5. One page or several. What it does: shows which searches fit on the same page. Limit: the AI does not see what already ranks on your site.

Look at the searches below and say which ones can be answered by the same page and which need separate pages.
For each group, explain why in one sentence and suggest the page title.

[paste the list of searches]

Which AI prompts help build an article brief?

They help build the structure, find gaps and list what needs a source. The AI builds the skeleton, but facts and your own examples remain yours.

6. Article brief. What it does: builds the goal, audience, sections and points that need a source. Limit: without seeing the search results, the structure comes out generic.

Build an article brief for the search "[keyword]".
Include: goal of the text, audience, search intent, a structure with section headings written as questions, points that need data or a source, and what the reader should do at the end.
Target length: [number] words.

7. Gaps against other pages. What it does: compares the section headings of three pages. Limit: paste only the headings, with no text from others.

Below are the section headings of three pages that already show up for the search "[keyword]".
List the topics all three cover, the ones only one covers and the ones none covers.
Do not copy sentences from the pages.

[paste only the section headings]

8. Claims that need a source. What it does: marks every number, date and study in the draft that needs proof. Limit: do not ask the AI for sources, since it can invent links.

Read the draft below and list every claim that needs a source: numbers, dates, studies, prices and tool names.
For each one, say what kind of source to look for (official documentation, study, the company's own data).
Do not suggest links or cite sources.

[paste the draft]

9. Internal interview script. What it does: writes questions for the company’s expert to give real examples. Limit: the AI asks, but the answers and numbers come from the person.

Create a script of 8 questions to interview [expert's job title] at our company about "[topic]".
The questions should draw out real examples, common mistakes and numbers that only people who do the work know.

10. Draft against brief. What it does: lists what the brief asked for and the text does not cover. Limit: it compares structure, not facts.

Compare the draft with the brief below.
List what the brief asked for and the text does not cover, what the text adds outside the brief and the answers that stayed vague.
Do not rewrite the text.

BRIEF: [paste]
DRAFT: [paste]

Which AI prompts for SEO help review the title and meta description?

They help generate options and find vague spots, but the decision is yours. Google sets no length limit for the title tag and truncates what does not fit the screen. In contrast, the meta description can be replaced by another part of the page.

11. Title tag options. What it does: writes ten titles that start with the keyword. Limit: the AI can get the character count wrong (our own caution), so check the preview in the plugin.

Write 10 title tag options for a page about "[keyword]".
Start each one with the keyword, describe the content accurately and avoid hype.
Then pick the 3 best and explain the choice in one line each.

12. Meta description options. What it does: writes five summaries with a call to click. Limit: Google may show another snippet.

Write 5 meta description options for the page below.
Each one should summarize what the person finds, use the keyword "[keyword]" once and end with a call to click, without promising results.

[paste the summary or the page text]

13. Review of the current title and meta. What it does: points out what is vague, repeated or different from the page. Limit: without click data, the AI cannot tell which one attracts more.

Evaluate the current title tag and meta description of this page.
Point out what is vague, repeated or different from the content and suggest a better version of each.

TITLE: [paste]
META DESCRIPTION: [paste]
PAGE: [paste the summary]

14. Repeated titles in bulk. What it does: rewrites only the repeated titles in a list of URLs. Limit: Google asks for distinct text on each page, but review them one by one.

The list below has the URL and title tag of pages on the site, and several titles repeat.
Rewrite only the repeated ones so that each page has a different title, true to the URL's topic.
Show a table: URL, current title and new title.

[paste the list]

15. Alt text. What it does: describes the image in up to 120 characters (our own cap, not Google’s). Limit: the AI only knows what you describe, so the alt must match the image.

Write the alt text for the image described below, in up to 120 characters.
Describe what appears, without starting with "image of" and without stacking keywords.

Image description: [describe]
Topic of the page where it appears: [topic]

How do you use AI prompts for technical SEO with exported data?

First, export the data as CSV (a plain-text spreadsheet). Then shrink the file and ask for the analysis in steps. AI summarizes and groups well, but exact math needs checking. Also, the steps and size limits are in the guide to analyzing Search Console with AI.

16. Crawl summary. What it does: counts the URLs with each issue, such as a missing title and a 404 error. Limit: the AI can get the count wrong, so ask for the math in code and check the total.

I attached "crawl.csv", exported from a tool that crawls the site.
Before analyzing, calculate in code and show the number of rows in the file.
Then count how many URLs have each issue: no title, repeated title, no meta description, 404 error and redirect.
Answer in a table with the count and three example URLs per issue, copied from the file.

17. Fix priority. What it does: ranks the issues by impact and effort. Limit: the AI does not know the business, so the final priority is the team’s.

Using the table of issues below, order the fixes by likely impact and effort, from highest to lowest impact.
Explain each position in one sentence and mark what depends on a developer.

[paste the table]

18. Target for 404 pages. What it does: suggests the closest live page for each error URL. Limit: the suggestion comes from topic, so open each target before you redirect.

Below are two lists: URLs with a 404 error and live URLs on the site, with the title of each one.
For each 404, suggest the closest live page by topic to receive the redirect.
Show a table: error URL, suggested target and reason. Mark the ones with no target.

[paste both lists]

19. Repeated titles and descriptions. What it does: groups the URLs that repeat the same text. Limit: in long lists, check the count in the file, because the AI can get it wrong.

In the file below, find the URLs that have the same title or the same meta description.
Group them by repeated text and say how many URLs are in each group.
Calculate in code and show the code.

[paste or attach the CSV]

20. Indexing report warning. What it does: explains the Search Console message and lists what to check. Limit: the AI cannot see your site, so the cause is a guess.

Search Console showed this warning in the indexing report: "[paste the warning text]".
Explain in plain language what it means and list what to check on the page, in order.
Say what you cannot know without seeing the site.

21. Issues into tasks. What it does: turns the list into tasks with an owner. Limit: deadlines are guesses, so agree on them with whoever does the work.

Turn the issues below into tasks.
For each one, say: what to do, who does it (content, developer or SEO), how to know it is solved and what depends on my decision.

[paste the list of issues]

Which prompts suit FAQs and structured data?

They help organize questions, draft answers with your own facts and build the structured data code, or schema. This code describes the page to Google and should reflect only what the page shows. Then validate it in Google’s Rich Results Test.

Also, Google stopped showing the FAQ result in search on May 7, 2026, according to the documentation updates log. Even so, an FAQ is still useful to the reader, and the guide to structured data for SEO shows what still counts.

22. Real questions for the FAQ. What it does: groups customer support questions and rewrites them the way people would type them. Limit: paste only questions with no personal data.

Below are questions that customers asked support, with no names or personal data.
Group similar questions, count how many are in each group and pick the 8 most frequent.
Rewrite each one the way a person would type it into Google.

[paste the questions]

23. FAQ answers with your facts. What it does: writes short answers using only the support text. Limit: the AI can add a fact, so compare each sentence with the source.

Write answers of 2 to 3 sentences for the questions below, using ONLY the facts in the support text.
If the support text does not answer, write "no data" instead of filling in.

QUESTIONS: [paste]
SUPPORT TEXT: [paste the confirmed facts]

24. Article JSON-LD. What it does: builds the code with the title, author, dates and image you provide. Limit: the code must match the page, and validity is checked in the tool.

Generate the Article JSON-LD for the page below, using only the data I provided.
If a field is missing, leave it out and list what was missing.

Title: [title]
Author: [name]
Publish date: [yyyy-mm-dd]
Update date: [yyyy-mm-dd]
Main image URL: [url]

25. Organization JSON-LD. What it does: builds the code with the name, logo and official profiles. Limit: the AI can make up fake profiles, so give only real URLs.

Generate the Organization JSON-LD with the data below.
Use only the profile URLs I provide in the sameAs list. Do not invent profiles or addresses.

Name: [name]
Site: [url]
Logo: [url]
Profiles: [URLs of the official profiles]

26. Code against visible text. What it does: points out fields the page does not show. Limit: it compares text, but the validity test is in the tool.

Compare the JSON-LD below with the visible text of the page.
List every field in the code that does not appear in the text, or that appears with a different value.

JSON-LD: [paste the code]
PAGE TEXT: [paste the text]

Which prompts help build the SEO report?

They help write and explain numbers you already calculated, not calculate in the dark. So compare each number in the text with the original table. Anonymize the client first, because reports often hold sensitive data.

27. Monthly summary. What it does: writes five lines from the month’s table. Limit: the AI can get the math wrong, so use only numbers you already calculated.

Write a 5-line summary for the client based on the table below.
Use only the numbers in the table, without recalculating or rounding. If a number looks odd, flag it instead of fixing it.

[paste the month's table]

28. What changed between two periods. What it does: describes what went up, down and stayed stable. Limit: cause is a guess, because the data shows what changed, not why.

The table has the same metrics in two periods.
Describe what went up, what went down and what stayed stable, with the values side by side.
Do not explain the causes. List separately three hypotheses I should check.

[paste both periods]

29. Next steps. What it does: proposes three to five actions with an owner and a way to measure. Limit: they are suggestions, so choose with the team.

Using the findings below, propose 3 to 5 next steps for the following month.
For each one, give the reason, the suggested owner and what will be measured to know if it worked.

[paste the findings]

30. Version for a non-expert client. What it does: explains the report without jargon. Limit: the AI can change a number when it rewrites, so compare all of them with the original.

Rewrite the technical report below for a client who does not know SEO.
Explain each term in one sentence, keep all numbers exactly as they are and do not add any new conclusion.

[paste the report]

What does Google say about using AI for SEO?

Google says generative AI can be useful for researching a topic and adding structure to original content. On the other hand, generating many pages with no value may break the policy on scaled content abuse. So these AI prompts for SEO are for research and review, not for mass pages.

In its guidance on generative AI, Google notes that models do not retrieve facts and can be inaccurate. For that reason, it asks:

“It is critical to manually factcheck and review all AI-generated content for accuracy and trustworthiness before publishing.”

Google Search Central (2026)

So the review also applies to the title, meta description, schema and alt text. The details are in the guide on whether Google penalizes AI content.

How do you check the AI answer before using it?

Compare each answer with a source you control, such as the exported file or the official page. AI prompts for SEO deliver the draft, and four checks make it safe.

  1. Numbers: recalculate totals and changes in the file, because the AI can get the math wrong.
  2. Facts and sources: open the original page for each data point, because the AI can invent information and sources in a confident tone.
  3. Promises: remove guarantees of results, such as “this will rank”.
  4. Metadata and code: check the preview in the SEO plugin and then validate the schema in the tool.

Finally, for monthly tasks, the guide to an AI agent for SEO shows what you can already delegate.

Want to know how your brand shows up in AI answers today?

Book a demo and see how your brand shows up in AI answers, with support from a GEO specialist.

Main sources

About this articleWritten with AI and reviewed by humans, with verified sources. Found something out of date? Write to the NotHuman team.

Get new NotHuman articles by email

By subscribing, you agree to receive blog articles by email and to the Privacy Policy. Unsubscribe any time.

Keep reading

Demo

Want to know how your brand shows up in AI answers today?

Book a demo and see how your brand shows up in AI answers, with support from a GEO specialist.

Book a demo

Your privacy

We use necessary cookies to make the site work and Google Analytics and Microsoft Clarity statistics cookies to improve our content. You can decline the statistics ones; in the EU, UK and Switzerland they are only turned on after you accept. Cookie Policy.

Cookie preferences

Cookie Policy