Why You Are Not Appearing in AI Overviews

You are not appearing in AI Overviews because Google is failing you at one of four layers: eligibility, retrieval, extractability, or authority. The visible answer may cite competitors that are not ranking number one organically, so checking a single keyword position will not identify the problem.

This is a Google-surface diagnosis, not a ChatGPT citation playbook. ChatGPT can draw from different retrieval systems, browsing behaviour, and source-selection patterns. AI Overviews are generated within Google Search and are constrained by the query, the available search results, the answer format Google needs, and its confidence in the sources it selects.

The diagnostic ladder for not appearing in AI Overviews

Do not start by rewriting every article or adding more “AI-friendly” language. Work through the layers in order. A page cannot become a cited source if it is ineligible to appear, cannot be retrieved for the query, cannot be cleanly extracted, or lacks enough authority for Google to rely on it.

Treat this as a ladder rather than a scorecard. Fixing authority will not overcome a noindex tag. Adding structured data will not solve poor retrieval. And a technically perfect page may still be omitted when Google does not need another source for that answer.

Layer 1: Confirm eligibility before analysing content

Eligibility is the technical minimum. Google cannot cite a page it cannot crawl, index, select canonically, or access in the version relevant to the searcher. This sounds basic, but it is routinely missed when teams audit only the URL shown in their CMS rather than the canonical URL Google has chosen.

Run these checks first

• Search the exact page title and a distinctive sentence in Google. If neither returns the intended URL, investigate indexing before content work.
• Inspect the URL in Google Search Console. Confirm that it is indexed, that Google selected the expected canonical, and that crawling is not blocked.
• Check for noindex, restrictive robots.txt rules, erroneous redirects, soft 404 behaviour, and login or consent walls that obstruct content.
• Compare desktop and mobile rendering. Important copy should appear in the rendered HTML, not only after a complex interaction.
• Check your target country and language variants. A US English page may not be the version Google selects for a UK query.

Use a 7- to 14-day window after a meaningful technical correction before deciding whether Google has processed it, although recrawling and re-evaluation can take longer on low-priority URLs. Do not submit the same URL repeatedly every day. That does not create a meaningful quality signal.

Also check page experience, but keep the diagnosis proportional. A page does not need a perfect lab score to be eligible. Google’s published Core Web Vitals guidance uses the 75th percentile of real-user data, with Largest Contentful Paint considered good at 2.5 seconds or less, Interaction to Next Paint at 200 milliseconds or less, and Cumulative Layout Shift at 0.1 or less. These are useful quality targets, not a guarantee of AI Overview inclusion.

Layer 2: Diagnose retrieval, not just keyword rankings

An AI Overview often answers several implicit questions at once. A query such as “best payroll software for a small UK business” may require eligibility criteria, pricing context, setup requirements, tax treatment, feature comparisons, and recommendations. Your page may rank for “payroll software” while failing to retrieve for the subtopic that triggers a citation.

For not appearing in AI Overviews, track the query cluster rather than one head term. Collect searches that produce an Overview, then record the claims in the answer and the cited URLs supporting each claim. You are looking for gaps between what Google is trying to answer and what your page demonstrably covers.

Build a retrieval map

1. Choose 20 to 50 commercially relevant queries where AI Overviews appear consistently in the location and device you serve.
2. Capture the Overview text, its cited domains, the organic results, and the date. AI Overview layouts can change, so dated snapshots matter.
3. Break each answer into atomic needs: definition, process, eligibility rule, price range, comparison point, risk, or recommendation criterion.
4. Map each need to your best existing URL. If no page owns the topic, decide whether to expand an existing page or create a focused resource.
5. Review Search Console query data monthly for 8 to 12 weeks. Look for growing impressions on the component queries, not merely movement on the headline term.

A practical editorial threshold is this: if a page is intended to support a multi-part query, it should answer at least 80% of the recurring subquestions you identify without forcing the reader to visit three other pages. That is an operating standard, not a Google rule. The remaining 20% may belong on specialised pages, provided internal links make the relationship clear.

Popular advice says that you must rank in the top 10 organic results to appear in an AI Overview. That is wrong as an absolute rule. Strong organic visibility helps because it makes retrieval more likely, but AI Overviews can cite sources beyond the most obvious first-page results. Conversely, a top-three ranking is not enough if the page lacks the specific evidence or format needed for the generated answer.

Layer 3: Make the answer extractable

Retrieval gets your page considered. Extractability makes it usable. Google needs to identify a passage that answers a specific point clearly, accurately, and with appropriate qualification. Dense thought-leadership prose, oversized product pitches, and claims hidden inside image assets create unnecessary friction.

Write for the question behind the query. Put the direct answer near the relevant heading, then provide explanation, exceptions, and proof. A useful answer block is often 40 to 80 words, but do not turn every section into a rigid snippet template. The goal is clarity, not a particular word count.

Use formats that reduce ambiguity

• Definitions: State what a term means, who it applies to, and any boundary condition.
• Processes: Use numbered steps when order matters, especially for setup, compliance, troubleshooting, and evaluation tasks.
• Comparisons: Use tables with consistent criteria such as price, implementation time, limits, integrations, and exclusions.
• Recommendations: State the decision rule. For example, identify when a solution is appropriate and when it is not.
• Evidence: Put the source, methodology, date, and caveat beside material claims rather than burying them at the bottom of a page.

Do not make unsupported superlatives easier to extract. “The best,” “leading,” and “most affordable” are weak unless you define the comparison set and evidence. If you use price ranges, state what they include. For example, “Plans typically cost $50 to $200 per month before implementation and usage fees” is more useful than “affordable pricing,” but only if you can support and maintain the range.

Structured data can help Google understand page entities and page type, but it is not an AI Overview switch. Do not add FAQ schema solely to chase visibility, and do not mark up content that users cannot see. Google’s rich-result policies require structured data to reflect visible page content. More importantly, schema cannot repair unclear writing or missing evidence.

Layer 4: Earn authority for the specific claim

Authority is not simply Domain Authority, a third-party metric rather than a Google ranking signal. For AI Overviews, the more useful question is whether your site and page appear credible for this exact claim. The bar rises for health, finance, legal, safety, and high-cost purchasing decisions because bad answers can cause real harm or loss.

Google’s Search Quality Rater Guidelines describe experience, expertise, authoritativeness, and trustworthiness as concepts used by human quality raters. They are not a checklist that guarantees rankings. Still, they offer a practical editorial test: can a reader establish who made the claim, why they are qualified, what evidence supports it, and when it was last checked?

Authority work that changes the underlying asset

• Publish original research with a stated sample, collection dates, methodology, limitations, and downloadable or inspectable supporting data where appropriate.
• Add named authors with relevant credentials and a clear review process for regulated or technical topics.
• Use first-hand examples: implementation screenshots, anonymised benchmarks, real workflows, test conditions, and documented outcomes.
• Build connected coverage around the topic, including definitions, alternatives, implementation, costs, limitations, and common failure modes.
• Earn relevant editorial mentions through genuinely useful data, tools, expert commentary, or research—not bulk guest-post placements.

Do not buy a package promising 100 backlinks for $99, publish 30 lightly rewritten articles in a week, or fabricate author personas. Those tactics may create activity, but they do not make a source more reliable. A credible original research project may cost roughly $2,000 to $15,000 or more depending on sample size, incentives, analysis, design, and promotion; a small expert-led guide can cost far less. The right investment depends on whether the topic has commercial value and whether you can sustain updates.

How to prioritise fixes over the next 90 days

Start with pages already close to visibility. A URL that receives impressions for relevant component queries and ranks between positions 8 and 30 is usually a better first candidate than a page with no indexation history and no topical relevance. This is a prioritisation rule, not a claim about a fixed Google threshold.

1. Days 1-14: Audit 20 to 50 target queries, validate eligibility, and identify the cited claim types competitors satisfy.
2. Days 15-45: Upgrade the highest-potential pages with direct answers, evidence, comparison structures, internal links, and clear ownership of subtopics.
3. Days 46-75: Create missing supporting assets where the retrieval map shows a genuine content gap. Avoid publishing near-duplicate pages for minor keyword variants.
4. Days 76-90: Add original evidence or expert review to the pages where trust is the limiting factor, then compare visibility and query coverage against the baseline.

Measure outcomes at three levels: AI Overview citation appearances for the tracked query set, organic impressions and clicks for component queries, and commercial behaviour such as qualified leads, demos, or revenue. A citation without useful traffic or pipeline is not automatically a win. Likewise, a page may improve business performance without becoming a visible citation on every query.

When absence is not a problem to fix

Some queries do not reliably show AI Overviews, and Google may vary the surface by location, device, intent, freshness, and search history. Others have an answer so short or so sensitive that Google may prefer a limited set of sources. You cannot force inclusion by publishing more words or by copying the exact phrasing of a cited competitor.

The objective is not to “rank in AI” as though it were a separate index. It is to become an eligible, retrievable, extractable, and credible source for the decisions your audience is making on Google. When the diagnostic ladder identifies the broken layer, the work becomes specific: fix access, close the retrieval gap, clarify the answer, or strengthen the evidence behind it.

Related reading

• AI search optimisation — scope, method and what we measure.
• Pricing
• The SEO ROI Formula, and the Inputs People Get Wrong
• Question Research Tools: Alternatives and Free Methods