Organic growth in the AI era: why AI visibility matters

Search engine optimisation has long helped people find products through pages, documentation and comparisons. It works alongside other sources of discovery, including recommendations, communities and partnerships.

AI assistants now provide another place to ask those questions. Their answers may name products, compare options and link to sources. The practical issue for a software company is how to measure that presence alongside the search and referral traffic it already tracks.

How AI search came into the picture

ChatGPT launched in November 2022, Perplexity a month later¹², Google switched on AI Overviews for every US user in May 2024¹⁰, and ChatGPT added web search that October¹¹. Under 4 years from first launch, the answers reach billions of people: 700 million weekly ChatGPT users by July 2025¹, and over 2.5 billion people a month seeing AI Overviews by mid 2026⁵.

Tap to zoom

OpenAI’s published usage study estimated ChatGPT’s reach at about 10% of the adult population by July 2025¹. Survey respondents and measured search visits are different populations: Pew’s March 2025 browsing panel found AI summaries in 18% of the observed Google searches³.

What people shopping for software do with AI assistants

Someone discovering software products from asking an AI assistant doesn't type keywords and scan a page of links like traditional keyword based search. Instead, they describe their situation and get back a generated summary, which may name a short list of products, with reasons, and sometimes links back to them. Multiply out OpenAI's usage figures and product questions alone come to an estimated ~55 million messages a day, in one assistant¹.

Surveys about software discovery differ in population and wording. They should not be combined into one adoption rate. The chart keeps each estimate with its question⁶⁷⁸. G2 reported that 33% of its respondents had bought from a vendor first suggested by a chatbot⁶; that is a result for its surveyed group.

Tap to zoom

What it does to clicks

Where an AI answer appears, people tend to click less. In Pew's browsing panel, a Google search with an AI summary ended in a click on a result 8% of the time, against 15% without one, and the summary's own sources were clicked on 1% of visits³.

Tap to zoom

Vendor panels agree on the direction and add the trend. Ahrefs measured the top result's click-through penalty deepening from 34.5% to 58% in ten months⁴, and Seer found the sites the answer doesn't cite losing the most¹³.

Tap to zoom

Google searches in Safari fell in April 2025 for the first time in 22 years, according to sworn testimony in the antitrust trial¹⁴. Google, for its part, keeps reporting record query volumes and growing search revenue⁵¹⁵. Both can be right: total search keeps growing while behaviour changes inside particular kinds of queries, and the queries with the most AI answers are still informational rather than commercial ones¹⁶. For now the assistants sit on top of search rather than replacing it. 95% of ChatGPT users still use Google¹⁷.

A citation can introduce a source to a reader without producing a visit. Seer’s panel also found differences in click-through rates between cited and uncited sites¹³. Those observations support measuring answer inclusion alongside traffic, but they do not establish that a citation was the final influence on a purchase.

How big it actually is

In Ahrefs’ mid-2025 study of 81,947 sites, AI sources accounted for 0.25% of the average site’s traffic¹⁸. A publisher panel reported under 1% of page-view referrals in early 2026¹⁹. These are separate samples and periods, rather than a universal share of web traffic.

Tap to zoom

Adobe reported a 1,200% rise in AI-sourced US retail visits over seven months, with a later update reporting 62% annual growth by July 2026²⁰. Reported conversion rates vary substantially: vendor accounts and a separate 78-site panel use different populations and baselines²¹²². They are reasons to inspect your own traffic, not conversion forecasts.

Website analytics can record referral visits and what happens after arrival. It cannot reconstruct all the recommendations a person read before clicking, or the answers in which a product was absent. Repeated answer checks provide another set of observations, with their own sampling limits.

What AI visibility is

AI visibility is whether the answer engines name your product, get it right, and point people to it when someone describes the problem you solve. The contested resource in the classic channel was a ranking; here it's what the answer says about you.

You'll see the work behind this sold under two acronyms, AEO (answer engine optimisation) and GEO (generative engine optimisation). In practice they're one family of work. We say AI visibility because it's describing the full scope of the challenge, but if you're interested, the full jargon table is on the explainer page, and AEO and GEO each have a deeper page of their own.

Our experiments on AI visibility so far

Our internal study asked customer questions for 152 companies and analysed the returned answers and citations. Company websites accounted for under 1% of citation instances in that corpus²³. That describes the linked sources; it does not measure every page retrieved or the effect of changing a company’s website.

Search pages, documentation and independent comparisons remain useful sources of information. Our citation count does not establish that publishing on a company’s own site has stopped working, or predict how future changes in web content will affect retrieval.

What this means for companies in the AI era

Start with a set of questions drawn from customer conversations. Record the assistant, search setting, date, full answer and sources. Repeat those questions over time, keeping errors separate from answers that omit the product. A stable record makes it possible to investigate changes without treating one response as a trend.

Inspect the answers before deciding what to change. Correct inaccurate information, explain a capability more clearly or fill a useful gap in a comparison. Keep website traffic and customer outcomes alongside those checks. AfterLaunch is the agentic growth engine; it monitors discoverability and prepares growth work with evidence and a draft.

Sources

1. 1Chatterji, Cunningham, Deming, Hitzig, Ong, Shan, Wadman (OpenAI, Duke, Harvard), NBER Working Paper 34255, September 2025.
2. 2Pew Research Center, "Americans and AI 2026", 17 June 2026.
3. 3Pew Research Center, measured browsing panel, 22 July 2025.
4. 4Ahrefs, AI Overviews CTR studies, April 2025 and February 2026.https://ahrefs.com/blog/ai-overviews-reduce-clicks-update/
5. 5Alphabet, Q2 2026 CEO remarks and June 2026 investor presentation.https://blog.google/alphabet/investor-presentation-june-2026/
6. 6G2, "The Answer Economy", n=1,076, released 15 April 2026.
7. 7G2, "2026 Buyer Behavior Report", a survey of more than 1,000 people who buy B2B software, released 22 July 2026.
8. 8Gartner survey (n=645 respondents, fielded Aug-Sep 2025), released 20 May 2026; Forrester blog, 22 January 2026; 6sense Buyer Experience Report, November 2025.
9. 9Stack Overflow, 2025 Developer Survey, 29 July 2025.
10. 10Google, AI Overviews US launch, 14 May 2024.
11. 11OpenAI, "Introducing ChatGPT search", 31 October 2024.
12. 12Perplexity company blog, March 2023 (recording the December 2022 launch).
13. 13Seer Interactive, AI Overview CTR panel, November 2025.
14. 14US District Court (D.C.), Memorandum Opinion, US v. Google remedies, September 2025 (Cue testimony in findings of fact).
15. 15Google, statement on search traffic, 7 May 2025.
16. 16Amsive, branded-query AI Overview study, April 2025.
17. 17Similarweb, "Is AI replacing search?", 4 August 2026; Bain snap chart, 20 March 2026.https://www.bain.com/insights/why-consumers-choose-an-ai-chatbot-over-a-search-engine-snap-chart/
18. 18Ahrefs, 81,947-site AI traffic study, June 2025.
19. 19Chartbeat publisher panel, as reported by Search Engine Journal, March 2026.
20. 20Adobe Analytics, March 2025; July 2026 update as reported by Digital Commerce 360, 19 August 2026.https://www.digitalcommerce360.com/2026/08/19/adobe-ai-referral-traffic-data-july-2026/
21. 21Ahrefs, June 2025; Vercel engineering blog, 10 June 2025; Semrush, July 2025. Each is a vendor reporting its own data.https://vercel.com/blog/how-were-adapting-seo-for-llms-and-ai-searchhttps://www.semrush.com/blog/ai-search-seo-traffic-study/
22. 22Siege Media, 78-site conversion panel, June 2026.
23. 23AfterLaunch, the 152-company corpus, August 2026.

Figures we deliberately did not use

• Gartner's "search engine volume will drop 25% by 2026" is a February 2024 prediction, not a measurement, and is routinely quoted as if it were one.
• Every "Safari searches fell X%" figure: no percentage exists on the court record.
• Perplexity query volumes dated after May 2025: the company has published none, and circulating figures are estimates without method.
• "AI referral traffic is about 1% of all web visits" and its vertical breakdowns: no identifiable primary publisher.
• "AI chatbots drive 54% of shortlist formation" attributed to G2: actively contradicted by G2's own July 2026 release.