AI Search Impressions Are a Vanity Metric

AI search impressions are about to become the new screenshot metric. A team opens Search Console, sees a shiny line going up, posts it in Slack, and decides the AI strategy is working. I think that is the wrong reaction. Google only rolled out its new Search Generative AI performance reports on June 3, and the first version is thin by design. It shows visibility. It does not show whether the mention came from a buyer question, whether the answer made your brand look credible, or whether anyone visited your site ready to buy. SEO is dead if all we do is swap one vanity graph for another.

Why AI search impressions feel bigger than they are

The excitement is easy to understand. Google finally gave SEO teams a native report for AI search visibility in Search Console. According to Google’s launch note, the report currently breaks data into only five dimensions: impressions, pages, countries, devices, and dates. That is useful. It is also a warning sign. If your dashboard only shows five top-line slices, you are not looking at buyer intent. You are looking at where your brand surfaced.

That matters because the first wave of AI-search software has trained people to celebrate being seen. Impression charts, mention counts, and share-of-voice screenshots feel concrete, especially for teams that spent years living inside rank trackers. But AI answers do not behave like blue links. One prompt can trigger several hidden fan-out searches. One mention can come from a throwaway comparison query with zero buying intent. One citation can make you look generic instead of trusted.

And the new report is still very young. Barry Schwartz noted that the available data appears to begin on May 18, 2026. That means many teams are about to build big narratives on top of a fresh and narrow dataset. Bad habit.

What AI search impressions cannot tell you

AI search impressions do not tell you whether the answer was any good for your business. That is the whole problem.

Search Engine Land, citing BrightEdge data, reported that AI-driven search referrals accounted for less than 1% of traffic from January to August 2025. Usage grew fast. Traffic quality did not magically become easy to measure. So if your team is cheering a spike in AI visibility while referral traffic stays flat and pipeline stays cold, the dashboard is not proof of success. It is a prompt to ask harder questions.

John Mueller made the same point in plainer language when he wrote, “You might not optimize for these if you focus too much on clicks instead of the overall value of your visits from Search.” That line matters because it cuts against the default SEO reflex. People love whatever number is easiest to export. AI search impressions are easier to export than trust, conversion quality, or downstream influence. That does not make them the right KPI.

They also cannot tell you which conversations shaped the answer. You do not know whether your appearance came from your homepage, a stale affiliate roundup, a Reddit thread where someone compared you fairly, or a Reddit thread where somebody trashed your onboarding. In AI answers, context is the product. The citation is just the receipt.

Why cited conversations matter more than AI search impressions

If you want to know whether you exist in AI answers, follow the cited conversation, not just the impression count.

That is where Reddit becomes hard to ignore. Search Engine Land cited analysis showing that 99% of Reddit citations in ChatGPT point to unique discussion threads, not subreddit homepages or polished brand profiles. That is a huge clue. AI systems are not rewarding the shell of a community. They are rewarding specific conversations with specific context.

Reddit also reported 126.8 million global daily active uniques in Q1 2026. That scale matters, but the bigger point is the shape of the content. Buyers ask messy questions there. Operators answer with scars, workarounds, and side-by-side comparisons. That is the stuff language models keep citing.

If you have already read our guide on how to rank in ChatGPT search, you know technical access still matters. Bing inclusion, crawlability, answer-first structure, and freshness are table stakes. But table stakes are not a moat. The moat is whether your brand shows up in the conversations AI systems trust when a buyer asks a real question.

That is also why our Reddit AEO playbook matters more than another visibility screenshot. You can buy a monitoring tool and still not exist in the cited layer. You can watch the graph go up and still lose the category because competitors own the threads buyers actually read.

What to measure instead of AI search impressions

I am not saying ignore AI search impressions. I am saying demote them. Put them in the same bucket as raw pageviews: interesting, directional, and dangerous when promoted to strategy.

Track four things instead.

First, track which cited URLs and conversations keep appearing for your category. If the same Reddit threads, comparison pages, and review sites show up over and over, that is your real search landscape.

Second, track qualified visits from AI surfaces. Not just volume. Look at bounce rate, assisted conversions, demo starts, signup quality, and whether visitors land on pages that actually answer the buying question.

Third, track sentiment and positioning inside cited third-party discussions. Did the answer frame you as the affordable option, the enterprise option, the risky option, or the tool people recommend only with caveats? Impression counts cannot answer that. Reading the conversation can.

Fourth, track whether your team is creating new citation surfaces. Helpful comments in relevant Reddit threads, sharper category pages, stronger comparison content, and updated explainers do more for GEO than another week of dashboard tourism. If you want a software layer for that workflow, our roundup of the best AEO and GEO tools is a better starting point than chasing one metric.

Here is the blunt version: AI search impressions will become popular because they are visible, native, and easy to paste into decks. They will also get abused for exactly those reasons. If your brand is getting seen but not trusted, cited, or chosen, you do not have momentum. You have decoration.

Google gave the market a new graph. Fine. But the teams that win GEO will be the ones that connect visibility to cited conversations and revenue, especially on Reddit, where the discussions are messy enough to sound human and specific enough to be useful. That is the wedge. That is the work.

So before you celebrate the next AI search impressions spike, ask a meaner question: where did the answer come from, and would that citation actually make a buyer trust you?