SEO is not dead. AI-assisted search still depends on many of the same foundations: crawlable pages, useful information, clear site structure, and evidence that a source is relevant and trustworthy.
Google’s current guidance says there are no special technical requirements for appearing in AI Overviews or AI Mode. A page must first be indexed and eligible to appear in normal Google Search. Google also says there is no special AI schema and no need to rewrite every page into tiny “AI-friendly” chunks. The official guide is unusually direct about this.
ChatGPT Search has a different crawler, but the first requirement is equally ordinary: a public site must allow OAI-SearchBot if it wants its content included in summaries and snippets. OpenAI documents this for publishers.
The useful work starts with the website itself.
1. Make important pages accessible
Check whether a crawler can reach the page, receive a successful response, and read the main content.
For each important page, verify:
- it is not blocked in
robots.txt; - it does not carry an accidental
noindexinstruction; - it has one canonical URL;
- the title and description match the page;
- the main information is visible without a login;
- internal links lead to it;
- the sitemap does not contain broken or redirected URLs.
JavaScript sites can be indexed, but they create more ways for metadata or content to fail silently. If a product, price, author, or article exists only after several browser requests, test what a crawler receives before those requests complete.
2. Publish information another site cannot copy
AI systems can summarize common advice from hundreds of pages. Repeating the same advice with different wording gives them little reason to use yours.
Useful source material includes:
- original measurements;
- a real case with dates and limits;
- a process you used;
- a comparison based on first-hand testing;
- a public dataset;
- a clear explanation from a person with direct experience.
Google now calls this non-commodity content: information based on real expertise or experience rather than another summary of common knowledge.
For nikmcfly.com, a report comparing streaming dashboards with royalty statements is more valuable than another article listing average per-stream rates. The report contains evidence that did not exist elsewhere before we published it.
3. Keep important facts clear and consistent
A visitor should be able to answer basic questions without combining several pages:
- Who created this?
- What does the company or product do?
- What result is being claimed?
- What exactly was measured?
- When was the information updated?
- Where can the original evidence be found?
Use the same name, product description, and canonical URL across your own important pages. Correct outdated profiles when possible. Do not manufacture third-party mentions or create profiles only to repeat a claim. Consistency is useful when it reflects reality; coordinated repetition is not independent proof.
4. Write pages for readers who scan
Clear structure helps people and machines locate the relevant part of a page.
A practical article usually needs:
- a descriptive title;
- the main fact or result near the beginning;
- section headings that state what each section contains;
- definitions for necessary terms;
- sources beside the claims they support;
- dates and limitations beside numbers;
- one clear next action.
This is not “chunking for AI.” It is ordinary editing. A paragraph should still make sense as part of the whole article, not as a pile of isolated search fragments.
5. Add structured data only when it is accurate
Structured data can identify an article, person, organization, product, event, or book in a machine-readable form. It can also make an error look official.
Add markup that matches visible information on the page. Test it. Update it when the page changes. Do not add unsupported awards, ratings, prices, or relationships.
Structured data may help a search system understand a page and may make the page eligible for some search features. It does not guarantee ranking or citation, and Google says it is not required for generative AI search.
6. Measure discovery and business results separately
Visibility is not one number.
Track at least three layers:
- Discovery: indexed pages, search impressions, and referrals from search or AI products.
- Citation: whether important questions produce an accurate mention and a working source link.
- Outcome: qualified visits, sign-ups, purchases, or inquiries from those sources.
Run a small, fixed set of real customer questions once a month. Save the answers, cited URLs, and date. Do not treat one personalized chatbot answer as a universal ranking.
Search Console can measure Google visibility. OpenAI says referrals from ChatGPT can be tracked in normal analytics. The two systems do not use the same reporting, and neither guarantees that a technically correct page will appear.
Start with five pages
Choose the five pages that best explain who you are, what you sell, and what you know from direct experience.
Make them crawlable. Remove contradictions. Add original facts and sources. Give each page a clear purpose. Then measure whether people and search systems can find the right answer.
That work is still SEO. AI search makes accuracy, original information, and technical clarity more important; it does not replace them with a new collection of hacks.
For a larger implementation sequence, see The Complete AIO Playbook.