A Top Position Can Still Be Practically Invisible
AI Overview position in Search Console can look better than the customer experience really is. A link inside an AI-generated answer may be reported with a top position because Google measures the answer block as a whole. That does not mean the customer saw your link, clicked it, trusted your business, or bought anything.
For an owner, the useful question is not “Did we rank number one inside the report?” It is “Did this visibility produce qualified visits, calls, bookings, orders, or evidence that buyers are considering us?” The first question earns a pleasant chart. The second one helps pay people.
Search Engine Journal recently described this reporting behavior as block flattening: Search Console treats an AI Overview as one search-result block rather than assigning a separate position to every link inside it (https://www.searchenginejournal.com/search-console-uses-block-flattening-for-aios-forget-position-focus-on-outcomes/589582/). A prominent citation and a less noticeable link can therefore receive the same reported position.
What Block Flattening Means
Google’s Search Console documentation says position is calculated relative to the topmost position occupied by a search-result element. All links and refinements inside the same element share that element’s position (https://support.google.com/webmasters/answer/7042828?hl=en). That rule matters when the element is a large, interactive AI Overview containing several sources.
Imagine an AI Overview appearing at the top of a results page. Your page is cited somewhere inside that block. Search Console can associate your link with the position of the whole block, even though the actual link might be immediately visible, lower in the answer, or available only after the user expands part of the experience.
That is block flattening in practical terms: several placements with different levels of human visibility can become one tidy position number. The report is not lying. It is answering a narrower question than many people assume.
Why Average Position Can Create False Confidence
Average position was already a blended metric before AI answers became common. Results can vary by query, device, location, date, and search layout. Combining those appearances into one average can conceal very different customer experiences.
A page could appear high inside an AI Overview for some searches and much lower in ordinary results for others. The average may look respectable while neither placement generates useful traffic. It is possible to win the spreadsheet and lose the customer, which is a remarkably efficient way to feel productive without making money.
Search Engine Journal also notes that impression rules can complicate the picture. A link loaded in a visible part of an AI Overview may count as an impression even when a person does not meaningfully notice it, while a link behind an interactive control may not count until that control is opened. Treat impressions and position as reporting signals, not eyewitness testimony.
Google’s own guidance says AI Overviews and AI Mode may use query fan-out, issuing multiple related searches across subtopics and data sources to develop a response (https://developers.google.com/search/docs/appearance/ai-features). That makes the experience even less like a fixed list of ten blue links. Different sources can support different parts of an answer, and the arrangement can change.
Use a Three-Layer Measurement Model
Owners do not need to throw away Search Console. They need to stop asking one metric to do every job.
Layer 1: Reported visibility
Use Search Console to monitor whether important pages appear, which queries and pages are involved, how impressions and clicks change, and whether performance shifts by device, country, date, or search type. Position can help identify a pattern, but it should not be the verdict.
Track:
- Impressions for commercially useful queries and pages
- Clicks and click-through rate
- Changes by device, location, and time period
- Landing pages gaining or losing exposure
- Branded versus non-branded demand where the data allows
Layer 2: Observed placement and message
For a controlled sample of valuable customer questions, inspect what a searcher actually encounters. Record whether an AI Overview appears, whether your business or page is cited, how prominent the source is, what claims the answer makes, which competitors are included, and whether your evidence is represented accurately.
Do not treat one manual search as universal truth. Search results vary. Use a repeatable prompt or query set, document dates and locations, and compare changes over time. The purpose is not to manufacture a perfect “AI rank.” It is to understand the customer-facing answer.
Layer 3: Business outcomes
Connect organic and AI-assisted visibility to actions that matter: qualified sessions, contact forms, calls, booked appointments, product views, purchases, directions, sales conversations, and revenue where attribution is credible.
A citation that produces no click can still influence awareness or trust, but do not invent value simply because direct attribution is difficult. Look for corroborating evidence such as branded search growth, customer-source responses, assisted conversions, or sales notes. Correlation is useful when labeled honestly. It is not causation wearing a nicer jacket.
Build an Owner-Level Scorecard
A practical monthly scorecard should fit on one page. Give each layer enough space to reveal a problem without turning the report into an archaeological dig.
Include:
- Visibility: important pages, query groups, impressions, clicks, CTR, and directional position changes
- Answer presence: sample questions tested, citations observed, source prominence, message accuracy, and competitor inclusion
- Customer behavior: engaged visits, calls, forms, bookings, purchases, and lead quality
- Revenue connection: qualified opportunities, assisted sales, closed revenue, and known attribution limits
- Next action: the one or two fixes most likely to improve customer acquisition or protect revenue
The scorecard should explain what changed, why it might matter, what the data cannot prove, and what the business will do next. If the report needs a 40-minute guided tour before an owner can identify the decision, it is not a report. It is a hostage situation with charts.
Diagnose a Strong Position With Weak Results
If Search Console reports a strong AI Overview position but clicks and customers remain weak, investigate the gap instead of celebrating the number.
First, inspect the actual answer experience. Your link may be present but visually buried, attached to a narrow factual claim, or less compelling than competing sources. Second, review the landing page. A citation can attract curiosity, but a vague page, stale offer, weak proof, slow load, or confusing next step can waste it.
Third, check whether the queries have business value. Informational visibility can build authority, but it should not be reported as if every impression were a sales opportunity. Separate research questions from comparison, local, product, pricing, and service intent.
Fourth, check whether your business is merely cited or meaningfully recommended. A source can support one sentence without placing the brand on the buyer’s shortlist. Recommendation readiness requires clear positioning, verifiable claims, useful service or product information, third-party proof, consistent entity details, and an easy conversion path.
Mistakes to Avoid
Do not sell an average position of 1 as proof of dominant AI visibility. The metric does not show exact visual prominence inside every AI answer.
Do not replace one vanity metric with another. A proprietary “AI score” still needs a transparent method, stable sampling, and a connection to customer behavior.
Do not make large website changes from one week of volatile data. Establish a baseline, segment the affected pages and queries, inspect the customer-facing experience, then prioritize changes closest to revenue.
Do not dismiss position entirely. It remains a useful directional clue when interpreted with impressions, clicks, page performance, observed layouts, and conversions. A hammer is useful. It simply should not be promoted to chief financial officer.
Measure the Customer Journey, Not the Trophy
AI Overview position in Search Console is useful when you understand its boundaries. Block flattening means a top reported position can describe membership in a top-level answer block, not guaranteed prominence, attention, or demand.
Keep the metric, but put it in its proper place. Pair reported visibility with observed answer placement and real business outcomes. That gives owners a more honest view of whether AI search is helping customers find, trust, and choose the business.
If the numbers look impressive but customers are not moving, an AI Visibility Audit can identify where visibility, message accuracy, proof, landing-page quality, and conversion paths stop connecting.
