To rank in AI Overviews, start with the basics Google itself points to: your page must be indexed and eligible to show a snippet, and it must be genuinely helpful. Beyond that, pages are easier to cite when they answer a specific question in the first sentences, use clear headings, state facts about their author or business consistently, and offer original experience.
I write my own articles this way now, and this post is part of my entity SEO series.
How do AI Overviews choose sources?
Google has not published a formula. Its documentation says AI Overviews draw on its core ranking systems and that the same SEO best practices apply. In practice, cited pages tend to be relevant, indexed, accessible to snippets, and contain a clear passage that answers the query. Being already visible in normal results for related queries helps.
Step 1: make sure you are eligible
- The page is indexed (check with URL Inspection in Search Console).
- No
nosnippet,max-snippet:0ordata-nosnippeton the content you want used. - Content is in the HTML or reliably rendered; not only in images or behind clicks.
- robots.txt allows Googlebot. Note that the separate
Google-Extendedtoken controls use of content for training Gemini models; Google states it does not affect inclusion in Search.
Many “AI visibility” problems are simply technical SEO problems.
Step 2: write answer-first
AI systems extract passages. Make good passages easy to find:
- Put the question as a heading: “What is a good LCP score?”
- Answer in the first 40 to 60 words under it, completely and plainly.
- Then add detail, examples and exceptions.
Compare:
“Many people wonder about LCP, and there are many factors to consider…”
with
“A good LCP is 2.5 seconds or less at the 75th percentile of real page loads.”
The second is quotable. The first is not.
Step 3: structure for extraction
- Logical H2/H3 hierarchy that reads like an outline.
- Short paragraphs, one idea each.
- Lists for steps and criteria; tables for comparisons.
- A summary near the top (I use a “key takeaways” box on my posts).
- FAQ sections answering related questions, with FAQPage markup where appropriate.
Step 4: cover the topic, not just a keyword
AI answers often combine several sub-questions. A topic cluster helps: a pillar page covering the topic broadly, supported by detailed pages on each sub-question, all linked together. This blog is organised that way; each category has a pillar such as the Core Web Vitals guide.
Use real questions people ask: Google’s autocomplete and “People also ask,” questions from clients, support emails and forums.
Step 5: show real experience
Generic content is easy to produce and therefore easy to ignore. Content that shows first-hand experience is harder to replace:
- Specific steps you actually follow.
- Settings, code and commands you have used.
- Mistakes you have seen and how you fixed them.
- Honest limitations and “it depends” explained.
Never invent statistics or case studies. If a fact can be checked, check it; if you cannot verify it, leave it out.
Step 6: make your entity clear
AI systems need to know who is speaking:
- Author name and role on each post, linked to an author or About page.
PersonorOrganizationschema withsameAslinks.- Consistent descriptions on all profiles.
This is the core of entity SEO. An optional llms.txt file can summarise the same facts for AI tools.
Step 7: keep content fresh and accurate
Update posts when facts change (for example when INP replaced FID), show a “last updated” date, and fix outdated advice. Stale content is a risk when it is quoted back to users.
Beyond Google: other AI assistants
Assistants that search the web live (and cite links) generally rely on a search index plus their own retrieval. To be eligible:
- Do not block the relevant crawlers in robots.txt unless you intend to. Each provider documents its user agents; decide deliberately which to allow.
- Be indexed in Bing as well as Google, since some assistants use Bing’s index.
- Keep pages fast and accessible to crawlers.
How to measure
There is no perfect “AI ranking” report. What I do:
- Track Search Console impressions and clicks for question-style queries.
- Regularly ask the actual questions in Google and AI assistants and note whether and how the site is cited.
- Check referral traffic from AI assistant domains in analytics.
- Watch branded searches over time; entity work shows up there.
What to avoid
- Hidden text or content written only for AI crawlers.
- Mass-produced pages with little value; Google’s spam policies cover scaled content abuse.
- Keyword stuffing question phrases.
- Fabricated statistics or reviews.
Checklist
- Indexed and snippet-eligible
- Question headings with direct 40–60 word answers
- Clear structure: summary, lists, tables, FAQ
- Topic cluster with internal links
- Original experience and verified facts
- Author entity clear, schema in place
- Crawlers allowed deliberately; Bing indexed
- Monitoring routine set up
Before and after: rewriting a section for extraction
Most pages do not need new topics. They need existing answers written so they can stand alone. An illustrative example from a web design agency’s pricing page:
Before: “Every project is different, and we believe in tailored solutions, so pricing depends on many factors that we discuss during our discovery call.”
After: “Website cost in Egypt depends mainly on three things: the number of unique page templates, whether you need e-commerce, and whether content is written in one language or two. A bilingual Arabic-English site usually costs more because each page needs separate copy and layout checks.”
The second version names the subject, gives the main factors and includes the local context. It can be quoted on its own without misleading anyone. It does not invent prices it cannot support.
Content types that tend to be useful for AI answers
- Definitions with context: what a term means and when it matters.
- Processes: numbered steps with what each step produces.
- Comparisons: tables that compare options on the same criteria.
- Decision guides: “choose X if…, choose Y if…” with honest trade-offs.
- Local specifics: payment methods, regulations, delivery norms or language needs in Egypt and the Gulf that generic global pages do not cover.
Google’s guidance on helpful, people-first content applies here directly. Nothing in this list is a trick; it is simply clear writing.
Arabic content and AI Overviews
Arabic searches often have fewer high-quality, well-structured pages than English ones, which is an opportunity for businesses in the region. A few practical points:
- Write the Arabic page as its own article in natural Modern Standard Arabic, not a literal translation.
- Use the terms people actually search, including English technical words where Arabic readers use them (SEO, WordPress, Schema).
- Link language versions correctly with hreflang so the right version is shown to the right reader.
- Keep facts identical across versions. If the English page says delivery takes 3 days and the Arabic page says 5, neither looks reliable.
A 30-day plan for a small site
- Week 1: confirm key pages are indexed in Search Console and not blocked by
nosnippetor robots rules. - Week 2: pick your five most important pages. Add a direct answer under each main heading and a short summary at the top.
- Week 3: add author information, update outdated facts and link related articles together.
- Week 4: record a baseline: Search Console clicks and impressions for those pages, and a manual check of which queries show an AI Overview and which sources it cites.
Repeat the manual check monthly. AI Overviews traffic is included in Search Console’s Web search data rather than shown separately, so trends on the improved pages are your best signal.
FAQ
Can I opt out of AI Overviews but stay in search?
Google documents preview controls such as nosnippet and max-snippet that also apply to AI features. They limit how your content is shown, which can reduce visibility in regular snippets too.
Do I need special markup for AI Overviews?
No. Google states that no special schema or file is required. Pages need to be indexed and eligible to show a snippet.
Why do AI Overviews cite a competitor instead of me?
Often their page answers the specific question more directly or covers a detail yours skips. Compare the cited passage with your page and fill the actual gap.
If you want your site and content set up for AI search visibility, see my services or contact me.
Frequently asked questions
Can I optimise specifically for AI Overviews?
Google's guidance is that normal SEO best practices apply and no special markup or files are required. Focus on helpful content that directly answers questions, on indexable pages that can show snippets.
Does nosnippet stop my content appearing in AI Overviews?
Google documents that nosnippet and related snippet controls also apply to AI features in Search, so using them limits how your content can be shown.
How do AI assistants like ChatGPT choose sources?
It depends on the product. Some answer from training data, others search the web live and cite results. Clear, well-structured, crawlable and widely corroborated content has the best chance in both cases.
Is AI search traffic tracked in Search Console?
Google includes traffic from AI features within the overall Web search performance data in Search Console rather than as a separate report.
