Blog

How to Use AI for SEO: A Complete, Step-by-Step Guide

The Coavon team · 18 October 2026 · 24 min read

To use AI for SEO, treat it as a fast, tireless assistant, not an author. Let AI do the slow parts of the job: brainstorming and grouping keywords, summarizing what already ranks, drafting outlines and briefs, writing first drafts of title tags and meta descriptions, generating schema markup, writing regular expressions and spreadsheet formulas, and spotting patterns in your Search Console data. Keep people on the parts that decide rankings: choosing what to target, adding first-hand experience and original information, checking every fact, and editing the final page. Then make sure the same pages are easy for AI search tools such as Google's AI Overviews and ChatGPT to find, read and cite.

The short version, as a workflow:

1. Brief the AI on your business once, so every answer starts from your audience, products and voice.
2. Research keywords with AI for ideas, clusters and intent, then check the numbers in a real keyword tool.
3. Study the results that already rank and ask AI what they all cover and what they miss.
4. Turn that into a content brief with an outline, questions to answer and the experience only you can add.
5. Draft with AI, then rewrite as a person: add examples, data and opinions, and check every claim.
6. Optimize the page: title tag, meta description, headings, internal links and structured data.
7. Refresh older pages that are slipping, using Search Console data.
8. Use AI for technical SEO jobs: regex, schema, redirects, robots.txt and spreadsheet work.
9. Make your pages easy for AI search to cite, and measure the traffic it sends.

The rest of this guide goes through each step in detail, with prompts you can copy, what Google says about AI content, the mistakes to avoid, and a weekly routine you can follow.

Jump to: what Google says about AI content, the step-by-step method, optimizing for AI Overviews and ChatGPT, or the prompt library.

What "AI for SEO" means

People use the phrase for two different things, and a good plan covers both:

• Using AI to do SEO work. Large language models such as ChatGPT, Claude and Gemini, and the AI features inside SEO tools, speed up research, writing, editing, analysis and technical tasks.
• Doing SEO for AI search. Search is no longer only ten blue links. Google shows AI Overviews and has an AI Mode, and people ask ChatGPT, Claude, Perplexity and others for recommendations. These tools read web pages and cite some of them. Being one of the pages they cite is part of SEO now, sometimes called generative engine optimization (GEO) or answer engine optimization (AEO).

The good news is that the two overlap. The pages that rank well in normal search, clear, accurate, useful and easy to crawl, are the same pages AI tools tend to pick up.

What Google says about AI-generated content

Before using AI to write anything, know the rules. Google's published guidance on AI-generated content says, in short, that it rewards helpful, reliable content made for people, however it was produced. Using AI or automation is not against its guidelines in itself. Using it mainly to manipulate search rankings is.

Google's spam policies name this directly as scaled content abuse: producing many pages mainly to rank, with little value for readers, whether they are written by people, by AI or by a mix. Sites that publish hundreds of thin, look-alike AI pages are exactly what that policy targets.

Google also asks creators to think about who made the content, how it was made, and why. Its quality guidelines put weight on experience, expertise, authoritativeness and trust (often shortened to E-E-A-T). AI can draft sentences, but it cannot have tried your product, run your experiment or served your customers. That first-hand experience is what you add.

Where AI helps in SEO, and where it does not

One rule runs through all of it: AI language models do not know your search data. Ask one for the monthly search volume of a keyword and it will often give you a confident number it made up. Use AI for thinking and writing, and use real tools (Google Search Console, Google Keyword Planner, or an SEO tool such as Ahrefs or Semrush) for numbers.

How to use AI for SEO, step by step

Step 1: Give the AI context about your business

Generic prompts get generic answers. Before any SEO work, write a short brief the AI can read every time: what you sell, who buys it, the problems they have, the words they use, your competitors, your tone of voice, and things you never say. Save it as a document, or in your AI tool's project or custom-instructions feature, and paste or attach it at the start of each piece of work.

Prompt: "Here is a brief about my business: [paste]. Ask me up to ten questions that would help you give better SEO and content advice for this business. Then rewrite the brief to include my answers."

Step 2: Do keyword research with AI

AI is excellent at the creative side of keyword research: thinking of every way a customer might describe a problem, and grouping the results. It is poor at the numbers. Use it in four passes:

1. Brainstorm seed topics. Ask for the problems, questions and tasks your customers search for, not just product terms.
2. Expand into long-tail keywords. Ask for questions, comparisons ("X vs Y"), "best" and "how to" forms, and terms used by beginners and experts.
3. Check the data. Put the list into a keyword tool to get real volumes, difficulty and the pages that rank. Drop terms nobody searches.
4. Cluster and label intent. Paste the checked list back and ask the AI to group keywords that one page could answer, and label each group's intent: informational, commercial, transactional or navigational.

Prompt: "Here are 120 keywords with their monthly search volume: [paste]. Group them into topic clusters, where every keyword in a cluster could be answered by the same page. For each cluster give: a suggested main keyword (the highest-volume one that matches the page), the search intent, the kind of page that should rank (guide, list, comparison, template, product page) and the total volume. Do not invent any keywords or numbers."

A useful habit: ask the AI to explain why two keywords belong on one page or two. If the reason is weak, look at the search results for both. If the same pages rank for both, one page can usually target them together.

Step 3: Analyze search intent and the pages that already rank

The pages on the first results page tell you what Google thinks searchers want. Open them yourself, then use AI to speed up the comparison. Copy the headings (or the full text) of the top five or so results and ask what they have in common and what is missing.

Prompt: "Below are the headings of the five pages that rank for [keyword]. 1) What intent do they all serve? 2) Which subtopics appear in at least three of them? 3) Which useful questions does none of them answer well? 4) What could a page add that none of them have, such as a tool, template, worked example or original data?"

Point 4 is the one that matters most. A page that only repeats what already ranks has no reason to outrank it.

Step 4: Turn the research into a content brief

A content brief tells the writer (human, AI, or both) exactly what the page must do. AI can produce a strong first version in a minute. A good brief contains:

• The main keyword, close variations, and the intent.
• Who the reader is and what they should be able to do after reading.
• A proposed title and H1, and an outline of H2 and H3 headings.
• The questions the page must answer, including ones from "People also ask" boxes and from customers.
• What only you can add: your own examples, numbers, screenshots, processes, mistakes and opinions.
• Internal pages to link to, and the call to action.
• Things to avoid: claims you cannot back up, competitor names you should not mention, words your brand does not use.

Prompt: "Using the analysis above and my business brief, write a content brief for a page targeting [keyword]. Include: reader, goal, title options, an H2/H3 outline, questions to answer, places where I should add first-hand experience (mark these [EXPERIENCE NEEDED]), internal links from this list [paste URLs], and a call to action."

Step 5: Write with AI, without publishing AI content

This is where most people go wrong. Asking an AI to "write a 2,000-word SEO article about X" produces a fluent, average page that says what every other page says, sometimes with errors. A better process puts the expert's knowledge in first:

1. Collect the raw material. Talk through the topic for ten minutes and record it, write rough bullet notes, or pull in an interview with an expert or a customer. Add your own data, screenshots and examples.
2. Let AI shape it. Give the AI the brief and your raw notes and ask for a draft that uses your material, in your voice, following the outline.
3. Rewrite as a person. Cut filler, sharpen the opening so it answers the question at once, add opinions and specifics, and remove anything that sounds like every other article.
4. Fact-check everything. Check every number, name, date, quote and claim against a source you trust. Delete what you cannot verify.
5. Have a second person review it. An editor or a subject expert reads it before it goes live.

Prompt: "Write a first draft for this brief: [paste]. Use only the facts and examples in my notes below; where you would need a fact I have not given you, write [CHECK: what is needed] instead of guessing. Answer the main question in the first two sentences. Short paragraphs, plain words, no filler phrases. My notes: [paste]"

The [CHECK] instruction matters. It turns the AI's tendency to fill gaps with invented details into a visible to-do list.

Signs a page still reads like raw AI output

• It opens with a sentence about how important or fast-changing the topic is, instead of the answer.
• Every section is the same length and ends with a tidy summary sentence.
• It uses vague claims ("many experts agree", "studies show") with no source.
• It has no examples, numbers, screenshots or opinions that could only have come from you.
• It could be moved to a competitor's site without changing a word.

Step 6: Optimize the page with AI

Once the content is right, AI makes the on-page details quick:

• Title tags. Ask for ten options under about 60 characters that include the main keyword near the start and give a reason to click. Pick one and edit it.
• Meta descriptions. Ask for options under about 155 characters that summarize the page honestly. Google may still write its own, but a good one helps.
• Headings. Ask the AI to list the headings and check that they follow a logical order, cover the brief and read well on their own.
• Readability. Ask it to flag long sentences, jargon and passive phrasing, then fix what you agree with.
• Image alt text. Describe each image and ask for short, specific alt text that would help someone using a screen reader. Do not stuff keywords into it.
• FAQs. Ask which follow-up questions a reader would still have, and answer the good ones on the page.

Prompt: "Here is my finished article: [paste]. Suggest 10 title tags (max 60 characters) and 5 meta descriptions (max 155 characters) for the keyword [keyword]. Then list any heading that does not match the section under it, and any question from the brief the article does not answer."

Step 7: Add structured data

Structured data (schema markup) helps search engines understand what a page is: an article, a product, a how-to, a recipe, a local business, an event. AI writes JSON-LD markup quickly and accurately enough to save a lot of time.

Prompt: "Write JSON-LD structured data for this page: an Article with headline [title], author [name], datePublished [date], publisher [organization and logo URL]. Use only properties from schema.org's Article type. Output only the code."

Always test the result with Google's Rich Results Test or the Schema Markup Validator before publishing, and only mark up what is actually visible on the page. Also note that Google has reduced some rich results over time; for example, FAQ rich results are now shown mainly for well-known government and health sites, so do not add FAQ markup expecting it to show for everyone.

Step 8: Build internal links with AI

Internal links help search engines find your pages and understand which ones matter most, and they keep readers moving. AI is good at matching pages to each other:

1. Export a list of your pages with their titles and main keyword (from your sitemap or a crawler).
2. Give the AI the new article and the list.
3. Ask for the five to ten most relevant pages to link to, the sentence each link should go in, and natural anchor text.
4. Ask the reverse question too: which existing pages should link to the new one.

Prompt: "Here is a list of my site's pages (URL, title, main keyword): [paste]. Here is a new article: [paste]. Suggest up to 8 internal links from the article to existing pages, quoting the sentence where each link fits and the anchor text. Then suggest 5 existing pages that should link to this new article and where."

Step 9: Refresh content that is slipping

Updating a page that already ranks on page one or two is often quicker than writing a new one. Use Google Search Console to find candidates: pages whose clicks have dropped compared with the same period last year, and pages with many impressions but a low click-through rate.

• Export the page's queries from Search Console and give them to the AI with the page text. Ask which queries the page does not answer well.
• Ask what is out of date: old years, old prices, old screenshots, tools that have changed.
• Compare the page with the results that now outrank it (Step 3), and add what is missing.
• Rewrite the title tag and meta description if the click-through rate is low.
• Update the visible date only when the content has really changed.

Prompt: "This page used to rank for [keyword] and has lost clicks. Here is the page: [paste]. Here are the queries it gets impressions for, with clicks, impressions and position: [paste]. List the queries it answers poorly or not at all, anything that looks out of date, and a prioritized list of changes."

Step 10: Use AI for technical SEO

You do not need to be a developer to handle many technical SEO tasks if you use AI carefully. Some of the most useful jobs:

• Regular expressions for Search Console. Search Console's filters accept regex. Ask the AI for one that matches, for example, all question queries: ^(who|what|why|how|when|where|can|does|is)\b. Test it in the filter and check the results make sense.
• robots.txt rules. Describe what to allow and block in plain words and ask for the rules, then test them with the robots.txt report in Search Console. A wrong line can block your whole site.
• Redirect maps. During a site move, give the AI the old and new URL lists and ask it to match them by topic. Review every match before it goes live.
• Spreadsheet formulas. Ask for formulas to merge exports, find duplicates, pull the domain out of a URL, or flag titles over 60 characters.
• Small scripts. Ask for a script that checks a list of URLs for their status codes, titles and canonical tags. Run it on a small list first.
• Page speed fixes. Paste a PageSpeed Insights report and ask the AI to explain each issue in plain words and suggest fixes in order of impact.
• hreflang and canonical tags for sites in several languages or with duplicate versions of pages.
• Log file analysis. Ask for help reading server logs to see which pages search engine crawlers visit, and how often.

Rule for all technical work: AI-written code and rules can look right and still be wrong. Test on a copy or a few pages first, and keep a way to undo the change.

Step 11: Use AI for link building and digital PR

Links from other reputable sites still matter. AI helps with the research and the writing, not with shortcuts:

• Find angles. Ask what data, tools or opinions you have that journalists or bloggers in your field would want to cite.
• Research prospects. Summarize a site's recent articles so your pitch refers to something real.
• Draft personal outreach. Write a short, specific first draft for each person, then rewrite it in your own words.
• Create linkable assets. Use AI to help analyze your own anonymized data into a report, or to plan a free template or calculator people will link to.

Avoid mass-produced AI outreach and AI-written guest posts placed only for links. Google's spam policies cover link schemes, and editors recognize template emails quickly.

Step 12: Use AI for local SEO

For businesses with a physical location or service area, AI can draft Google Business Profile posts and descriptions, suggest service and category wording, and help write replies to reviews. Keep review replies personal and specific, never post fake reviews (they break Google's policies and, in many places, the law), and write a real, distinct page for each location you actually serve rather than near-copies with the town name swapped.

Step 13: Report and analyze with AI

Search Console and analytics exports are where AI saves the most time for analysis. Export the data (queries, pages, clicks, impressions, click-through rate, position) and ask questions about it:

• Which queries have high impressions and a low click-through rate? (Candidates for better titles.)
• Which pages rank in positions 5 to 20 for valuable queries? (Candidates for improvement.)
• Which queries does more than one of my pages appear for? (Possible cannibalization.)
• What changed between this month and last? Which pages or queries explain most of the change?
• Summarize this month's results in five bullet points for my manager.

AI tools with data analysis can work through a spreadsheet directly, and many can write the charts for you. Check a few of its numbers against the original export: AI can misread columns or totals.

How to optimize for AI search: AI Overviews, ChatGPT and more

Optimizing for AI search builds on normal SEO rather than replacing it. Google says there is nothing special you need to do to appear in AI Overviews or AI Mode beyond the usual SEO basics: the page must be indexed and eligible to show a snippet in Search. Other AI tools work in a similar way, searching the web and citing pages they can read. These are the steps that help most.

1. Make sure AI search tools can reach your pages

Each AI company uses named crawlers, and some separate the crawler that trains models from the one that powers search answers. You can allow one and block the other in robots.txt. Check each company's own documentation, because names change. Commonly used ones:

Blocking Googlebot removes you from Google Search entirely, so never do that to stay out of AI Overviews. Also make sure your important content is in the page's HTML, not only drawn in by JavaScript after the page loads: some AI crawlers do not run JavaScript.

2. Answer the question early and clearly

AI answers are built from passages that answer a question directly. Put a clear, complete answer in the first sentences under each heading, then add detail. Use descriptive headings that match the questions people ask, lists for steps, and tables for comparisons.

3. Be specific and verifiable

Exact facts with a clear source are easier to cite than vague claims: real prices with the date you checked them, step-by-step instructions, definitions, specifications, and original data you collected. Original research and data are among the best ways to be cited, because no one else has them.

4. Build your brand beyond your own site

AI tools often recommend brands that are discussed widely: in reviews, comparisons, forums, news and lists on other sites. Real mentions come from the same work as good PR: being useful, being quoted, publishing data, appearing on podcasts and in industry lists, and earning reviews from real customers. Keep your company's name, description and key facts consistent everywhere.

5. Keep facts about your business correct and current

If AI tools describe your product wrongly, check what your own pages say first: old pricing pages, outdated feature lists and contradictory descriptions get repeated. A clear About page, an up-to-date pricing page and honest comparison pages give AI tools accurate material to quote.

6. Measure AI search traffic

• Google: clicks from AI Overviews and AI Mode are counted in Search Console's Performance report as part of Web search, not shown separately.
• ChatGPT, Perplexity, Claude, Copilot and Gemini: look in your analytics for visits referred from their domains. ChatGPT also often adds utm_source=chatgpt.com to links.
• Ask the tools yourself. Once a month, ask several AI assistants the questions your customers ask, and note whether you are mentioned, how you are described and which pages are cited.

About llms.txt: it is a proposed file that lists a site's key pages for AI tools. It is easy to add, but the major search and AI companies have not said their search products rely on it, so treat it as optional and do not expect it to change your visibility on its own.

Types of AI tools for SEO

You do not need many tools. Most teams use one general AI assistant plus the SEO data tools they already have.

Connecting your AI assistant to the place your work lives saves the most copying and pasting. For example, if your content plan, briefs and drafts are in Coavon, the Coavon connector lets Claude read and create them directly (see below).

A prompt library for SEO

Copy these, replace the brackets, and always attach your business brief.

• Topic ideas: "List 40 questions [audience] asks before buying [product type], grouped by stage: just learning, comparing options, ready to buy."
• Keyword expansion: "For the topic [topic], list long-tail searches a beginner and an expert would type, including questions, comparisons and 'best' searches. Do not include search volumes."
• Intent check: "For each keyword below, say whether the searcher wants to learn, compare, buy or find a specific site, and what kind of page would satisfy them: [paste]."
• Content gaps: "Here are our blog's titles: [paste]. Here are a competitor's: [paste]. Which topics do they cover that we do not, that matter to [audience]?"
• Outline: "Create an H2/H3 outline for [keyword] that answers the question in the first section, covers [subtopics], and ends with an FAQ of 6 real questions."
• Intro rewrite: "Rewrite this introduction so the first two sentences directly answer [question], with no throat-clearing: [paste]."
• Fact list: "List every factual claim, number, name and date in this article so I can check each one: [paste]."
• Simplify: "Rewrite this section for a reader new to the topic, keeping every fact: [paste]."
• Title tags: "Write 10 title tags under 60 characters for [keyword] for this page: [summary]."
• Meta descriptions: "Write 5 honest meta descriptions under 155 characters for this page: [summary]."
• Alt text: "Write concise alt text for an image showing [description], for a page about [topic]."
• Schema: "Write JSON-LD for a [type] with these details: [details]. Only use schema.org properties."
• Regex: "Write a Search Console regex that matches queries containing [words] but not [words]."
• Search Console analysis: "From this export, list queries with over [n] impressions, CTR under 2% and an average position under 10, and suggest a better title for each page: [paste]."
• Cannibalization: "From this export of query and page pairs, find queries where two or more of our pages get impressions, and suggest which page should own each query: [paste]."
• Refresh plan: "Compare our page with these headings from pages that now outrank it, and list what we should add, cut or update: [paste both]."
• Outreach: "Write a 90-word email to [name] at [site], referring to their article [title], offering [our data or resource]. Plain and personal, no flattery."
• AI visibility check: "What are the best [product category] tools for [audience]? Cite your sources." (Ask this in several AI tools each month and record the answers.)

Common mistakes when using AI for SEO

• Publishing AI drafts without real editing. Fluent is not the same as useful, and scaled thin content is what Google's spam policies target.
• Trusting AI with numbers. Search volumes, statistics, prices and dates from a language model are often invented. Verify every one.
• Inventing experience. Fake case studies, quotes, reviews or author bios damage trust and can break laws on deceptive advertising.
• Producing look-alike pages at scale, such as hundreds of city or keyword pages with only a word changed.
• Copying the top results. If a page only rephrases what already ranks, there is no reason for it to rank higher.
• Ignoring the brief. Without context, AI writes for an average reader of an average business.
• Shipping untested technical changes. A wrong robots.txt line, redirect or canonical tag can remove pages from search.
• Blocking the wrong crawler. Blocking Googlebot to avoid AI Overviews also removes you from Google Search.
• Pasting private data into tools. Check your AI tool's data settings before sharing customer data, and remove personal details first.
• Measuring only rankings. Track clicks, conversions and AI referrals too, since AI answers can change how many people click through.

A weekly AI SEO routine

For a fuller checklist of the technical side, use our free SEO audit template. To plan what to publish and when, the content calendar template and our guide to content marketing strategies are good next reads.

Running an AI SEO workflow in Coavon

An AI SEO process has a lot of moving parts: keyword lists, briefs, drafts, reviews, publishing dates and refreshes. Coavon keeps them in one place, and its connector lets Claude and other AI assistants work with them directly.

• Plan in [Flow](/product/flow). Put each article on a board with custom statuses such as Brief, Drafting, Review and Published (on Pro), assign a writer and a due date, and set recurring tasks for the weekly review and monthly refreshes.
• Write in Docs. Keep briefs and drafts in Docs, where editors leave comments and suggested edits, and download a finished piece as Word or PDF if needed.
• Require sign-off. Flow's approvals let you name an approver, such as an editor or a client, and a task cannot be marked done until it is approved.
• Keep research in Notes, your private notebook for keyword ideas and interview notes.
• Collect requests with a form. A public Flow form turns each content request into a task.
• Connect Claude. With the Coavon connector, you can ask Claude to read your brief, search your notes, create a draft document in Docs, and create or update tasks in Flow, so the AI works inside your process rather than in a separate chat window.

You can try it free; see pricing for what each plan includes.

Frequently asked questions

How do I use AI for SEO?

Use AI for the slow, repetitive parts of SEO: brainstorming and clustering keywords, summarizing the pages that rank, writing content briefs and first drafts, suggesting title tags and meta descriptions, writing schema markup and regex, and analyzing Search Console exports. Keep people responsible for choosing targets, adding first-hand experience, checking facts and editing the final page, and check every number in a real SEO tool.

Is AI-generated content bad for SEO?

Not in itself. Google says it rewards helpful, reliable content made for people however it is produced, and that using AI is not against its guidelines. Using AI mainly to manipulate rankings, such as publishing large numbers of thin pages, falls under its spam policy on scaled content abuse.

Can Google detect AI content?

Google focuses on whether content is helpful and original, not on whether AI was involved. Pages that are thin, repetitive or made mainly to rank can perform badly whoever or whatever wrote them, so edit AI drafts heavily and add your own experience and information.

Can AI do keyword research?

AI is good at generating keyword ideas, long-tail variations and questions, grouping keywords into topics and labelling search intent. It does not have reliable search volume or difficulty data, so check the numbers in Google Keyword Planner, Search Console or an SEO tool such as Ahrefs or Semrush.

How do I get my website into Google AI Overviews?

Google says no special optimization is needed beyond normal SEO: your page must be indexed and eligible to show a snippet. Help your chances by answering questions clearly near the top of each section, using descriptive headings, being specific and accurate, and keeping important content in the HTML.

How do I get my business mentioned by ChatGPT and other AI assistants?

Make sure their search crawlers (such as OAI-SearchBot for ChatGPT) can reach your site, publish clear and accurate pages about what you offer, keep your facts consistent across the web, and earn genuine mentions in reviews, comparisons, news and industry lists. Check monthly by asking the assistants the questions your customers ask.

What is the best AI tool for SEO?

Most teams do well with one general AI assistant, such as ChatGPT, Claude or Gemini, combined with Google Search Console and one SEO data tool for real keyword and backlink data. Content optimization tools and crawlers are useful additions for larger sites.

Will AI replace SEO?

AI is changing how people search and how SEO work is done, but the core job remains: making useful, trustworthy pages that search engines and AI tools can find, understand and recommend. AI makes that work faster; it does not remove the need for expertise and judgement.

What is GEO or AEO?

Generative engine optimization (GEO) and answer engine optimization (AEO) are names for optimizing content so AI tools such as Google AI Overviews, ChatGPT and Perplexity cite or recommend it. In practice it builds on SEO: crawlable pages, clear answers, accurate specifics, original information and a well-known brand.

Coavon · Blog · Templates · Features · Pricing