CLIENT REVENUE GENERATED$9,720,200+Independent consulting ยท Since 2008
AI SEARCH DIAGNOSIS
You can find your page in Google, but an AI answer cites another website. I would not rush to rewrite the page on that evidence alone. First, I want to know which answer you tested and whether the page was a useful source for that particular question.
I would work through three questions: can the page qualify, does it answer what the buyer needs to know, and what did the test actually show? Those checks help you choose a sensible next step instead of reacting to every answer that leaves you out.
Indexing is a starting condition, not a promise of selection. Google says a page needs to be indexed and eligible for a search snippet to qualify as a supporting link in AI Overviews or AI Mode. Meeting those conditions does not guarantee inclusion. Google’s AI features guidance also says there is no special AI file or schema required for these Search features.
Those are Google’s requirements. Do not assume that Google indexing proves access or eligibility in ChatGPT, Perplexity or another answer product. Investigate the product you actually tested.
My starting point would be the observation, then eligibility, then the page’s usefulness for the question. I would only recommend a change once there is a specific finding to address.
“We are not showing in AI” is too broad to investigate. Name the product and experience: an AI Overview in Google Search is not the same observation as an AI Mode response or an answer in a separate assistant.
Keep a small record for each check:
If no AI Overview appears, record that as a different event from an Overview that appears but does not cite you. Likewise, a business mention without a link is not a linked citation. Do not combine these into one unexplained success or failure score.
For a repeatable log and reporting method, use the existing AI search visibility measurement guide. This article focuses on deciding what to investigate after observing absence.
Inspect the specific URL in Search Console rather than relying on a general site search. Check its recorded indexing status and Google-selected canonical. A live test and the indexed record answer different questions: being fetchable now does not establish that the indexed version has been updated.
A duplicate URL may not be the version Google selected. Review redirects, canonical signals, sitemap entries and internal links together before editing them. Do not remove a valid canonical merely because a preferred page was absent from an answer. See Google’s canonical URL guidance and our explanation of crawling versus indexation.
Check both the page’s robots meta tag and HTTP X-Robots-Tag headers. For example, nosnippet restricts previews; data-nosnippet applies to marked portions of content. Other preview limits need to be interpreted according to their documented behaviour, not removed wholesale. Google documents these controls.
Before removing a restriction, I would ask who put it there and why. Privacy, licensing and contractual requirements matter more than clearing an SEO warning. Get the responsible owner’s approval before changing an intentional control.
A previously indexed page can later develop a server, firewall or rendering problem. Inspect the rendered page and required assets, then review actual failed requests where evidence is available. Do not declare a crawler blocked because an answer omitted you, and do not identify a genuine crawler solely from its User-Agent name.
Keep administrator and private routes protected. Public-content diagnosis is not a reason to disable security across the website.
Once access and eligibility checks pass, move from technical troubleshooting to the actual buyer’s need. A page about your general expertise may not help someone comparing a particular service, delivery area or implementation constraint.
For example, a question about support for a specific ecommerce platform needs more than a broad “we improve visibility” statement. The useful check is whether the relevant page explains the actual supported platform, work involved, limitations and credible evidence.
Use search intent to assess task match and internal linking to assess discovery. Neither is a guaranteed citation mechanism.
Inspect the cited pages as a comparison exercise: what question do they answer, and what evidence is visible? Their selection does not reveal the complete reason they were selected. Copying their headings, length or schema is not a demonstrated fix.
These are possible conditions to investigate, not confirmed causes for your site. Copy the rows into your working backlog and add the affected URL, owner and evidence date.
| Observed condition | Evidence to collect | Proportionate next action |
|---|---|---|
| Indexed and snippet-eligible, but not cited | Exact question, product, mode, answer and sources across repeat checks | Review task fit. Do not assume a technical fault from absence alone. |
| A different canonical is selected | Inspected URLs, canonical tags, redirects, sitemap and linking signals | Reconcile genuinely conflicting signals; preserve intentional consolidation. |
| Preview restrictions affect the page | Robots meta, response headers, marked content and the restriction’s purpose | Ask the owner whether the restriction should change. Do not bypass privacy decisions. |
| The page only partly answers the buyer’s task | Question requirements compared with actual visible page content | Add missing, accurate detail to the appropriate page, not keyword padding. |
| Public business facts conflict | First-party facts and the specific conflicting profiles or source pages | Correct the authoritative records and request legitimate third-party corrections. |
| Observation coverage is too limited | Scheduled versus completed checks, missing answers and changed test conditions | Improve the sample and record limitations before claiming a trend. |
Illustrative scenario only: this is not a client result or a recorded platform experiment.
A hypothetical retailer asks, “Who can help with a WooCommerce site migration while preserving existing search traffic?” A hypothetical provider’s service page is indexed and snippet-eligible but is not cited in the observed answer.
The page says it offers ecommerce SEO, yet does not describe whether migration work is available, which platform it supports or how URL changes are reviewed. That is a useful content gap to investigate, not proof of the cause of omission. My AI content strategy work can turn those missing explanations into a scoped content plan.
The content improvement is useful even if the next answer still does not cite the page. Do not invent a service, promise traffic preservation or label a later citation as proof that this single change caused it. Once a gap is verified, AI Search optimisation can be scoped around the specific improvements.
Write down the finding, evidence, affected URL, responsible owner and completion test. A verified access error has a different next action from a missing service detail. When the evidence only shows absence, the next action may be better observation rather than a rewrite.
After an approved change, verify the page, links, rendering and relevant controls. Record the deployment date. Repeat the question without silently changing the market, mode or prompt. Keep unrelated changes in the log so a later movement is not automatically attributed to your edit.
If the eligibility checks pass and the page answers the question well, I would stop making reactive edits. Changing the headline every time an answer changes makes it harder to learn anything. Agree a review window and check whether the page helps suitable prospects, not just whether it was cited once.
If you need to investigate a wider set of questions, competitors, sources and platforms, a scoped AI visibility audit is an appropriate diagnostic next step. It should identify observations and priorities, not promise a citation.
No. It is not a promise that the page will be selected for any particular answer. Check the relevant experience and distinguish eligibility from observed inclusion.
Google says these are not required for AI Overviews or AI Mode. Accurate structured data can still describe appropriate content, but it is not a citation guarantee. Other products need their own current guidance.
No. Check access through actual fetch or request evidence. A missing answer alone does not establish a firewall problem.
Only when you can explain the gap and the benefit of the change. Fix verified errors or missing useful information; do not rewrite solely in response to an isolated absence.
There is no reliable deadline or inclusion promise. Verify the change, retain comparable observations and separate page quality, answer appearances and business outcomes.
Platform requirements checked on 5 October 2026. The diagnostic worksheet and scenario are editorial methods and illustrations, not published experimental findings.
