AI for SEO: a practical workflow that stays on Google's good side

AI for SEO works best as an assistant for research, planning and technical chores, with people supplying the expertise, original insight and final review. This guide covers what Google's guidance says about AI content, then walks through keyword clustering, briefs, drafting, on-page work, internal links, schema and AI Overviews, with prompts you can reuse.

How can you use AI for SEO?

Use AI for SEO to speed up the research, planning and technical work around content, and keep people responsible for expertise, originality and accuracy. AI is strong at sorting, summarizing, drafting and writing code. It is weak at knowing real search data and at having first-hand experience.

Here is where AI earns its place in an SEO workflow, and where it needs a human:

The rest of this guide goes through each step. First, the rules.

What does Google say about AI-generated content?

Google's position is that it judges content by quality and usefulness, not by whether AI helped write it, but using AI to mass-produce low-value pages violates its spam policies. Three Google documents spell this out.

The spam policies define scaled content abuse as many pages generated for the primary purpose of manipulating search rankings, and list "using generative AI tools or other similar tools to generate many pages without adding value for users" as an example.

Google's guidance on using generative AI content says AI can be useful for researching a topic and adding structure to original content. It also asks you to fact-check and review AI output before publishing, explicitly including title elements, meta descriptions, structured data and image alt text, because those can appear in search results. For ecommerce, AI-generated images should carry IPTC DigitalSourceType metadata set to TrainedAlgorithmicMedia.

The helpful content guide frames it as "Who, How, and Why":

• Who: use accurate bylines from real people, linked to background on the author. Invented personas are a low-quality signal.
• How: consider explaining how automation was used when readers would reasonably expect it.
• Why: the most important question. Content should exist mainly to help people, not to rank.

How do you do keyword research with AI?

Do keyword research with AI by giving it real query data and asking it to cluster, label intent and find gaps. Do not ask it for search volumes: language models do not have live search data and will invent plausible numbers.

1. Export real data. From Google Search Console, export queries with impressions, clicks and average position for the last 3 to 6 months. Add a keyword list from your research tool if you use one.
2. Ask the AI to clean and cluster. Group queries that one page could answer, and label each cluster with intent.
3. Map clusters to pages. Mark which clusters an existing page already covers and which need new content.
4. Prioritize by impressions, current position and business value, not by how easy the AI thinks a topic is.

Below is a CSV export from Google Search Console: query, clicks, impressions, CTR, position.1. Remove branded queries containing "acme".2. Group the remaining queries into clusters that a single page could fully answer.3. For each cluster, give: a short cluster name, search intent (informational, commercial,   transactional, navigational), total impressions, best current position, and the 5 top queries.4. Flag clusters with high impressions and position worse than 10 as opportunities.Return a markdown table. Do not invent any numbers that are not in the data.{{paste CSV}}

For larger exports, a spreadsheet or a short script does the arithmetic more reliably than a chat window. The guide to AI for data analysis covers how to check AI-produced numbers.

How do you write SEO content briefs with AI?

Write briefs with AI by giving it the target query, the pages that currently rank (which you read yourself), and what your business knows that others do not. Ask for the questions a reader needs answered, not for keyword density.

A useful brief contains the primary query and intent, the reader and what they want to do next, the questions to answer in order, entities and terms a thorough answer mentions, what top results miss, your unique material (data, screenshots, a process, a case), and internal pages to link.

Create a content brief for the query "how to choose a standing desk".Reader: office worker buying their first standing desk, budget under $600.I read the top 5 results; they all list features but none explain how to measure desk heightfor your own body. We sell desks and have return data on the top reasons people send them back.Give: search intent, a suggested H2 outline phrased as reader questions, the specific facts eachsection needs (mark any you are unsure of with [VERIFY]), and where our return data fits.

The [VERIFY] marker matters. It makes the AI separate what it knows from what it is guessing, and gives your editor a checklist. If you want to understand why models state wrong facts with confidence, read AI hallucinations.

Can AI write content that ranks?

AI can produce competent first drafts, but drafts that rank and keep ranking need the things only you have: first-hand experience, original data, specific examples and a clear point of view. Without them, an AI draft reads like a summary of the pages that already rank, which gives Google little reason to prefer it.

A workflow that keeps quality high:

1. Human sets the angle in the brief: what this page says that others do not.
2. AI drafts section by section from the brief, not the whole article from one line.
3. Human adds experience, screenshots, numbers, quotes from your own team, and cuts generic filler.
4. Fact-check every claim, number, product detail and link against a primary source.
5. Edit for voice. Remove the phrases that make text read as machine-written. The guide on making AI writing sound human has an editing checklist.
6. Real byline from the person responsible for the content.

Answer-first structure helps both readers and search features: start each section with a one or two sentence answer, then the detail. This makes sections easy to quote and easy to skim.

How do you use AI for on-page SEO and internal links?

Use AI to generate options for titles, meta descriptions, headings and internal links, then choose and edit them yourself. Generating twenty variants is fast; picking the right one still needs judgment.

Titles and meta descriptions

Write 8 title tags (max 60 characters) and 5 meta descriptions (140 to 160 characters) for this page.Primary query: "how to choose a standing desk". Put the query near the start of the title.Each description must say what the reader gets. No clickbait, no claims not in the page.Count characters and show the count next to each.{{paste page intro and H2s}}

AI models count characters poorly, so check lengths with a tool or script before publishing.

Internal links

Give the AI a list of your URLs with their titles (from your sitemap or a crawl) and the article text, then ask it to suggest three to six links with natural anchor text and the sentence each belongs in. Reject suggestions that only match a keyword without helping the reader. Internal links also help Google find pages: its AI features guide lists making content findable through internal links as a best practice.

Refreshing old pages

Paste an older article and its Search Console queries, and ask which queries the page gets impressions for but does not answer well. That list is often the fastest win in SEO.

How do you generate schema markup with AI?

AI writes valid JSON-LD structured data quickly, but you must validate it and make sure it matches what the page visibly shows. Google's generative AI guidance specifically calls out structured data as something to review before publishing.

Write JSON-LD Article markup for this page. Use only facts visible on the page:headline, author name and URL, datePublished, dateModified, image URL, publisher.Do not add fields I have not given you.

A typical result looks like this:

{  "@context": "https://schema.org",  "@type": "Article",  "headline": "How to choose a standing desk",  "author": { "@type": "Person", "name": "Jane Doe", "url": "https://example.com/team/jane" },  "datePublished": "2026-10-01",  "dateModified": "2026-10-15",  "image": "https://example.com/images/standing-desk.jpg",  "publisher": { "@type": "Organization", "name": "Example Desks" }}

Test the markup with Google's Rich Results Test before shipping. Also check which rich result types Google still shows: per the Search Central documentation updates, FAQ rich results stopped appearing in Google Search in May 2026, and several other structured data types were retired in 2025. Unused markup does no harm, but it will not earn a rich result either.

How does AI help with technical SEO?

AI helps most with technical SEO by writing the small scripts, regular expressions and queries that technical work depends on, and by explaining crawl or log data. It turns a 30-minute lookup into a one-minute task, as long as you test what it gives you.

Common uses:

• Regex for Search Console filters, such as all queries that are questions or all URLs under a folder.
• Redirect maps during a migration: give old and new URL lists and ask for a mapping, then review the uncertain ones.
• Crawl export analysis: find pages with missing titles, duplicate descriptions, broken internal links or thin content.
• Log file analysis: summarize which sections Googlebot crawls most and least.
• Fixing the site itself: a coding agent can update templates, add canonical tags, compress images or fix internal links across a codebase in one reviewed pull request.

For the last one, treat the agent like any developer: a clear task, tests or checks, and a review before merging.

Do you need to optimize for AI Overviews and AI search?

You do not need special markup or files to appear in Google's AI Overviews or AI Mode. Google's AI features documentation states there are no additional requirements: a page must be indexed and eligible to show with a snippet, and it must meet the normal technical requirements. Google also clarified in June 2026 that llms.txt files are not needed for Google Search.

What helps, according to that page, is ordinary good practice: allow crawling, keep important content in text, link pages internally, provide a good page experience, support text with quality images or video, and keep structured data consistent with visible content. If you want to limit what appears, the usual controls apply: nosnippet, data-nosnippet, max-snippet and noindex.

Other AI assistants that cite web pages reward the same things: clear, direct, well-sourced answers that can be quoted on their own.

AI for SEO checklist: start this week

1. Export the last 3 months of Search Console queries and have AI cluster them by intent.
2. Pick the three clusters with high impressions and weak positions.
3. Write a brief for each with AI, including what only you can add.
4. Draft with AI section by section, then add your experience and data.
5. Fact-check every claim and every AI-written title, description and alt text.
6. Generate and validate JSON-LD, and add three to six internal links.
7. Publish under a real byline, then track the queries in Search Console for 4 to 8 weeks.

For a wider view of AI across campaigns, ads and email, see AI for marketing. If your technical SEO backlog lives in a code repository, a coding agent on an always-on box such as Orkio can work through it as reviewed pull requests while you do other things.

Frequently asked questions