I Checked My Own AI Search Data Before Writing Another AEO Article

After publishing so much about AI and SEO, I wanted to see what Google’s own data said about my site.I’ve written quite a few articles about AI tools, SEO experiments, content quality, and the changing search landscape.I’ve compared AI assistants, tested prompts, explored AI agents, written about…

After publishing so much about AI and SEO, I wanted to see what Google’s own data said about my site.I’ve written quite a few articles about AI tools, SEO experiments, content quality, and the changing search landscape.I’ve compared AI assistants, tested prompts, explored AI agents, written about SEO campaigns, and examined how AI is changing the way marketers create and distribute content.So when Google introduced its new Generative AI performance report in Search Console, I did what I’d normally advise a client to do:I checked my own data first.Not a prediction. Not a screenshot from someone else’s account. Not another article explaining what AI search might do to SEO.My own site. And the first thing I learned was that Google’s new report answers one question better than I expected:Are my pages appearing inside Google’s AI search experiences?It doesn’t yet answer the question most publishers want to ask:How many people clicked?Inside Google’s New AI ReportGoogle’s Search Console Generative AI performance report provides visibility into impressions from its AI-powered search features.The report brings together three surfaces:AI OverviewsAI ModeAI Overviews in DiscoverYou can examine the data by page, country, device, and date. At this stage, Google reports impressions but doesn’t yet provide clicks, click-through rate, average position, or the exact queries that triggered an appearance.Google’s AI-search report currently shows visibility data, but not complete traffic attribution.That distinction matters.An impression means that a page appeared in an AI search experience. It doesn’t prove that someone read the citation, clicked through to the website, remembered the brand, or became a customer.The report isn’t a traditional traffic report. It’s better understood as a visibility and citation report.My Digitalboxes DataWhen I opened the report for Digitalboxes, I expected the results to follow a familiar pattern.I assumed the pages with the strongest traditional SEO performance would also receive the most AI-search impressions. I also expected the report to make it easy to understand which AI surface was generating the visibility.It didn’t.The report gave me useful information, but not a complete explanation. I could see which pages were appearing, yet I couldn’t see the exact prompts, queries, or individual AI surfaces behind those impressions.That created an interesting gap.Google was showing me where my site appeared, but it wasn’t fully explaining why it appeared.This is important because SEO professionals are used to working with query-level data. We analyse the search term, position, page, click-through rate, and conversion path.The generative AI report removes much of that context.It gives us the result before giving us the full reason.I compared Digitalboxes’ Generative AI feature impressions across two consecutive 28-day periods. The site recorded 4.34K impressions in the most recent period, compared with 4.24K in the previous period — an increase of approximately 100 impressions, or 2.4%.Digitalboxes recorded 4.34K Generative AI feature impressions in the most recent 28-day period, compared with 4.24K in the previous period.It isn’t a dramatic jump, but it does suggest that AI-search visibility remained relatively stable and moved slightly upward.Digitalboxes recorded 4.34K Generative AI feature impressions in the most recent 28-day period, compared with 4.24K in the previous period.The Pages Were the Real InsightThe most useful part of the report wasn’t simply the total number of impressions. It was having a baseline I can monitor over time.The comparison showed that Digitalboxes’ AI-search visibility was relatively stable across the two periods. However, the report still doesn’t provide enough detail to explain which queries or individual AI surfaces produced those impressions.Seeing the Pages dimension also gives me a way to evaluate my content over time. Instead of asking only which pages rank in traditional search, I can also look at which pages appear in Google’s AI-generated answers.That doesn’t mean Google is declaring those pages perfect. It does suggest that they contain something its systems found relevant, clear, or useful for a particular search experience. This creates an opportunity for deeper analysis.I can compare the pages that receive AI-search impressions with pages that don’t. I can look for differences in:How directly the page answers the main question.How clearly the information is structured.Whether the opening section provides a useful answer quickly.Whether the page includes original insights or practical examples.How well the content matches the likely search intent.The report doesn’t tell me exactly what to change. But it gives me a better starting point for investigating what may already be working.The Result I Didn’t ExpectThe surprising part wasn’t a single page or query. It was how little the report explains behind the overall number.Digitalboxes recorded 4.34K impressions in the most recent 28-day period, compared with 4.24K in the previous period. That small increase gave me a useful baseline, but it didn’t tell me which queries, pages, or AI surfaces were responsible for the change.For me, that raised a more useful question:What makes one page easier for an AI system to use than another?I don’t think there’s a single answer. It may involve structure, clarity, topical relevance, original information, or how directly the page addresses a question.That’s why I’m cautious about turning one report into a universal AEO formula.A Different Way to Think About AEOThe report changed how I think about AEO.I don’t see it as a separate trick that replaces SEO. I see it as another way to evaluate whether a page is clear, useful, specific, and easy to understand.The question is no longer only:Can this page rank?It’s also:Can this page help answer a question clearly enough to be selected as a source?That doesn’t mean every paragraph needs to be written for a machine. It means the content should make its answers easier to find and interpret.Clear headings, concise explanations, relevant examples, original research, and first-hand experience can help both readers and search systems understand the page.The best AEO advice I’ve seen so far is still closely related to good SEO and good editorial practice.Why Impressions Still MatterImpressions are an imperfect metric, but they aren’t meaningless.For the first time, site owners can see whether their pages are appearing inside Google’s generative AI features instead of relying entirely on manual searches or third-party tracking tools.That gives us a new layer of evidence.A decline in clicks doesn’t automatically mean that a site has disappeared from search. Some pages may still be appearing in AI answers, even when fewer users visit the source website.The opposite is also possible. A page may receive AI-search impressions without producing measurable visits. That’s why I wouldn’t interpret the report in isolation.I’d compare it with:Organic clicks in Search Console.Branded search growth.Direct traffic.Assisted conversions.Newsletter registrations.Leads and sales from returning visitors.The report shows exposure. Other analytics platforms help determine whether that exposure produces business value.The Citation EffectThe commercial value of appearing in an AI answer is still being studied, but early research suggests that citations can matter.The source I used for this analysis cites an estimate that brands included in AI Overviews received approximately 35% more organic clicks than brands that weren’t cited.I’d treat that figure as an external benchmark, not a promise or a result that can be applied directly to Digitalboxes.It doesn’t mean every citation will produce a click. It also doesn’t mean that impressions can be treated as revenue.But it does suggest that there may be a meaningful difference between being mentioned as a source and being invisible inside the answer.The click may be less common than it was in traditional search. However, when a user does decide to visit a source, the citation may provide an important trust signal.In that sense, AI search isn’t simply removing websites from the customer journey. It may be changing the point at which a website earns attention.The Gaps in the DataThe current report has several limitations.First, it doesn’t provide click or click-through-rate data. That means you can’t calculate the direct traffic contribution of AI Overviews or AI Mode from this report alone.Second, the AI surfaces are blended. You can’t clearly separate the performance of AI Overviews, AI Mode, and AI Overviews in Discover in the current view.Third, there’s no query-level information. You can see which pages appeared, but not the precise questions or prompts that caused them to appear.That makes it difficult to connect a specific content decision with a specific AI-search outcome.The report also needs to be interpreted carefully because Search Console data has experienced reporting issues during the rollout of these AI-related reports. A sudden drop in the figures shouldn’t automatically be treated as a real visibility or traffic decline.For now, I’d use the report to identify patterns and trends rather than present it as a complete attribution system.My Next StepI’m not planning to rewrite every page simply because it did or didn’t receive AI-search impressions. That would be an overreaction.Instead, I plan to use the report as a monthly content research tool. I’ll:Record which Digitalboxes pages receive AI-search impressions.Compare those pages with their traditional organic clicks.Review the structure and opening sections of pages that appear frequently.Look for patterns in clarity, direct answers, original information, and topical focus.Improve related pages without forcing unnatural AEO language into them.Track branded searches, direct traffic, and conversions alongside AI impressions.I also want to compare the pages appearing in the report with the pages I would’ve selected myself. That difference may be more valuable than the impression total.If Google’s systems repeatedly choose a particular page, that page may reveal something about how the content is structured or how directly it addresses user intent.The goal isn’t to copy one page across the entire website. The goal is to understand the qualities that make that page useful and apply those lessons thoughtfully to other content.The Lesson for My Content StrategyAfter publishing so much about AI and SEO, I’ve become increasingly cautious about confident predictions.There’s a lot of advice about “optimising for AI search,” but much of it is still based on limited observation, small datasets, or assumptions about how different systems work.Google’s report doesn’t solve that problem completely. What it does provide is a starting point for testing those assumptions against real site data.For me, the practical lesson is simple:Don’t create content only to satisfy a search engine or an AI system.Create pages that make the answer easy to find, easy to understand, and worth trusting.That means using:Clear structure.Specific explanations.First-hand experience.Original insights.Useful examples.Accurate and up-to-date information.Those are good practices for traditional search, human readers, and AI-generated answers at the same time.Check Your Own Report FirstIf the report is available in your Search Console account, check your own pages before reading another generic AEO checklist.Look at which URLs appear. Compare them with your assumptions. Notice what’s missing. The most useful insight may not be the total number of impressions. It may be the page you didn’t expect to see, or the page you expected to see but didn’t.Google has started showing us whether our pages are entering AI search experiences. Now we need to become better at interpreting what that visibility actually means.For me, the report isn’t a replacement for analytics, conversion tracking, or editorial judgment. It’s an early signal.And early signals are most useful when you compare them with reality instead of treating them as the whole story.This story is published under the Generative AI publication. Connect with us on LinkedIn and follow Zeniteq to stay in the loop with the latest AI stories. Let’s shape the future of AI together!I Checked My Own AI Search Data Before Writing Another AEO Article was originally published in Generative AI on Medium, where people are continuing the conversation by highlighting and responding to this story.

Source: Generative AI Pub — Published September 7, 2026 at 02:09 UTC — Category: Image AI