Marketing

Getting Found When Search Answers the Question Itself

Abstract cover art for the article: Getting Found When Search Answers the Question Itself

AI answers are taking a real share of the clicks that used to reach your website, and no formatting trick brings them back. What still works is older and harder. Publish material worth citing, make it obvious who you are, build demand for your name, and keep your pages readable by the systems that summarize them. Below is what the evidence shows, where it runs out, and where to put the effort.

What the click data says

The cleanest independent study so far comes from Pew Research Center. In July 2025 it published an analysis of the browsing activity of 900 US adults during March 2025. Across 68,879 Google searches, 18% produced an AI summary. When a summary appeared, people clicked a traditional result in 8% of visits. Without one, they clicked in 15%. Links inside the summary itself got clicked in 1% of visits. People were also more likely to end their browsing session after a page with a summary: 26% of the time, against 16% for pages with only traditional results.

That study describes ordinary people doing ordinary searches, which is both its strength and its limit. The detail B2B marketers should notice is which queries triggered summaries. Pew found that 60% of searches beginning with a question word such as “what” or “why” produced one. Question-shaped queries are exactly what most company blogs are built to catch.

Companies with access to aggregated Search Console data point the same way, with larger effects. Ahrefs compared 300,000 keywords and reported in February 2026 that a top-ranking page earned a 58% lower desktop click-through rate when an AI Overview was present, comparing December 2025 with December 2023. Seer Interactive, tracking queries for 42 client organizations, found that organic click-through on queries with an AI Overview fell from 1.76% in June 2024 to 0.61% in September 2025.

Google disputes the framing. In an August 2025 post, Liz Reid, who runs Search, wrote that total organic click volume to websites had been “relatively stable year-over-year” and that the clicks sent were of higher quality. The post included no figures for either claim. Read it as Google’s position, stated without published data.

What nobody knows yet

The direction is clear. The size, and whether it lasts, is not. Seer’s April 2026 update, drawn from a larger set of 53 brands, showed click-through on AI Overview queries climbing from 1.3% in December 2025 to 2.4% in February 2026. Its authors warned against forecasting a recovery from two months of data, which is the right instinct.

Measurement is still thin. Google’s generative AI performance report in Search Console, rolled out to all sites as of August 31, 2026, shows impressions in AI Overviews and AI Mode. It does not show clicks. You can now see that you appeared. Google still won’t tell you what appearing was worth.

Assistants are more opaque again. There is no public, independent equivalent of the Pew study for ChatGPT or Perplexity, and analytics referral data misses the buyer who reads an answer, remembers a name, and types your URL a week later. Anyone quoting a precise share of “AI traffic” for your category is estimating.

Assistants are a second front door

Scale is not in question. Sam Altman said in October 2025 that more than 800 million people use ChatGPT every week. For software purchases specifically, G2 surveyed 1,076 B2B decision-makers in March 2026 and found that 51% now start their research with an AI chatbot more often than with Google, up from 29% in G2’s previous report. G2 runs a review marketplace and benefits from that conclusion, so discount it a little. It still matches how most founders now shop for tools themselves.

The practical consequence is that part of your shortlisting happens in a conversation you never see. The assistant describes you using whatever it can find: your site, review platforms, forums, press, and the comparison page a competitor wrote about you. Your website is one input among several. It is usually the only one you fully control, so it should be the clearest.

What to do about it

Write pages that deserve to be cited

Google’s guidance here is unusually blunt. Its documentation on AI features says there are “no additional requirements” to appear in AI Overviews or AI Mode. Its optimization guide, last updated in July 2026, separates commodity content, the generic list of tips, from “unique expert or experienced takes that go beyond common knowledge.” The same guide says you don’t need llms.txt files, content chopped into small chunks, or special schema to show up in generative results.

For a B2B company, citable usually means specific. A page that states your pricing model, the integrations you support, who the product suits and who it doesn’t, and how implementation works gives an answer engine something concrete to quote. A post titled “Why onboarding matters” gives it nothing. Put the answer in the first paragraph, name things precisely, and keep every claim checkable.

The tradeoff is maintenance. Specific pages go stale faster and need an owner. If nobody will keep your pricing or comparison page current, a wrong one does more damage than a missing one, because the assistant will repeat it with confidence.

Make it obvious who you are

Answer engines summarize entities: companies, products, people. If your company shares its name with three others, your product was renamed last year, or your homepage describes you in abstractions, the summary will be vague or wrong. State what the company does in plain words on the homepage and the about page. Use the same name, one-line description and founding facts everywhere you control, including LinkedIn, Crunchbase, app marketplaces and review profiles. Organization structured data is reasonable housekeeping, but Google says structured data isn’t required for its generative features, so don’t expect it to move anything by itself.

Publish something only you have

If your product generates data, aggregated and anonymized findings from it are the most defensible content you can make, because nobody else can write them. Benchmarks, annual usage reports and honest analyses of patterns across your customer base all qualify. So does a well-argued point of view from someone who has done the work, the kind of “thoughtful first-person analysis” Reid’s post says is doing well.

This is slower than commissioning another explainer. It also compounds. Original data gets picked up by journalists and analysts, and those pieces become sources the assistants read. Clear anything drawn from customer data with legal first, and publish the sample size and method on the page. A number without a method is the claim a careful reader discounts, whether that reader is a person or a model.

Build demand for your name

When the answer box absorbs generic questions, branded search becomes the part of your traffic you can still rely on. Someone who types your company name has already decided to look for you. That demand mostly comes from outside SEO: a product people recommend, founders who write and speak in public, press, partnerships. Seer’s 2025 data showed brands cited in AI Overviews getting 35% higher organic click-through than uncited ones, and the researchers were careful to say they can’t prove citation causes the lift. Stronger brands may simply get cited more often. Either reading points to the same work.

Get the technical basics right

None of this matters if the systems can’t read your pages. Each AI product runs its own crawler, and blocking the wrong one by accident is common, often after a security or CDN change adds blanket bot rules that nobody reviewed with marketing. These are the ones most B2B sites should check in robots.txt and at the firewall.

Crawler Operator What blocking it does
Googlebot Google Removes pages from Search, including AI Overviews and AI Mode. For finer control, Google points to nosnippet, max-snippet and noindex.
Google-Extended Google Opts out of Gemini training and grounding. Google says it does not affect inclusion or ranking in Search.
OAI-SearchBot OpenAI Pages won’t be shown in ChatGPT search answers, though they can still appear as navigational links, per OpenAI’s crawler documentation.
GPTBot OpenAI Opts out of training OpenAI’s models. It is separate from search visibility.
PerplexityBot Perplexity Pages won’t be surfaced and linked in Perplexity search, per Perplexity’s documentation.

The distinction matters if you want content kept out of model training but visible in answers. Those are separate switches, and plenty of sites have flipped both without meaning to. Beyond crawler rules, the ordinary list still applies: main content present in the HTML, sensible internal links, fast pages, and a sitemap that matches the site you actually have.

When this matters less

If most of your pipeline comes from outbound, referrals, or a sales team working a named account list, AI search is a secondary concern. Fix the crawler rules and the homepage description, then get back to what already works. The same is true if your buyers mostly search for you by name.

The exposure sits with companies that built demand on informational content, the “what is X” and “how to choose Y” posts. That is the traffic every study above shows shrinking. Before rewriting your content plan around it, check in your CRM how much closed revenue that traffic actually produced. For some companies the answer is a lot. For many it was mostly top-of-funnel volume that looked good in a dashboard.

A next step for this quarter

Write down the twenty questions a buyer asks before choosing a product in your category. Ask each one in ChatGPT, Perplexity and Google. Note whether you appear, what is said about you, and which sources get cited. Then open the generative AI report in Search Console and list which of your pages already earn impressions there. The gap between those two lists is your work order for the next few months. Run the same check again next quarter, because the answers will have moved.

Sources