How Buyers Find Developers Through AI Search, and What Gets You Named

Referral sessions from AI assistants convert at several times the rate of Google organic sessions on the same site. Seer Interactive published a case study in 2025 measuring this on one business-to-business client, covering data from October 2024 to April 2025. | Traffic source | Conversion rate | |---|---| | ChatGPT | 15.9% | | Perplexity | 10.5% | | Claude | 5.0% | | Gemini | 3.0% | | Google organic (baseline) | 1.76% | Seer's own framing was that "AI traffic is converting at a much higher rate than Google Organic, with each LLM converting at a higher clip". The spread between the assistants is interesting on its own. ChatGPT referrals converted at roughly nine times the organic baseline; Gemini referrals converted at less than twice it. Two things to hold onto before you act on this table. It is one client site in one industry, so the absolute numbers are not a benchmark you can apply to your own business. And a conversion rate tells you nothing about how many people arrived. That second point is the one that gets left out of almost every blog post quoting these figures, and it changes what you should do about them.

In the same Seer dataset, AI assistants sent roughly 11,000 referral sessions against roughly 14,000,000 Google organic sessions. AI referrals were about 0.07% of organic volume. A conversion rate of 15.9% on 11,000 sessions is a smaller number of deals than 1.76% on 14 million. This is the whole shape of the opportunity as at mid-2026. Very high intent, very low volume. The person who arrives from an AI assistant has usually already had a conversation about their problem, been given a shortlist, and picked you off it. They arrive pre-qualified. There are just not many of them yet. Seer's data supports the pre-qualified reading. ChatGPT referral sessions averaged 2.3 pages per session against 1.2 for Google organic, with a 62% engagement rate. People arriving from an AI conversation read more of the site before they act. The practical conclusion is that AI search is worth a small, deliberate amount of effort rather than a rebuild of your marketing. Treat it the way you would treat a referral partner who sends you three good leads a year: worth keeping warm, not worth restructuring the business around.

The most-cited sources in AI answers are reference sites, community platforms and media, not the websites of the businesses being discussed. Profound analysed 680 million citations between August 2024 and June 2025 across ChatGPT, Google AI Overviews and Perplexity, and published the share of all citations going to each domain. | Engine | Most-cited domains by share of all citations | |---|---| | ChatGPT | Wikipedia 7.8%, Reddit 1.8%, Forbes 1.1%, G2 1.1% | | Google AI Overviews | Reddit 2.2%, YouTube 1.9%, Quora 1.5%, LinkedIn 1.3% | | Perplexity | Reddit 6.6%, YouTube 2.0%, Gartner 1.0%, Yelp 0.8% | Not one vendor website appears in any of those lists. The engines are reading pages written by third parties and summarising what those pages say about businesses. For a service business this is uncomfortable but clarifying. Rewriting your homepage does not change what Reddit, an industry directory or a review site says about you, and those are the pages doing the work. Existing on other people's pages is the mechanism. That means real case studies published by clients, genuine participation in communities where your buyers already are, accurate listings in the directories your category uses, and being written about by people who are not you.

Unlinked mentions of your business name across the web correlate more strongly with visibility in Google AI Overviews than backlinks do. Ahrefs studied 75,000 brands in August 2025 and found a correlation of 0.664 between branded web mentions and AI Overview brand visibility, against 0.218 for backlinks. That is roughly three times the strength for a signal most businesses do not track at all. Ahrefs are explicit that correlation is not causation, and so are we: the study does not prove that getting mentioned causes you to appear, and there are obvious confounders, since larger and better-known businesses accumulate both mentions and visibility for the same underlying reasons. The directional lesson survives the caveat, though. Ten years of search advice trained service businesses to chase links. The measurable signal that lines up best with being named by an AI answer is being talked about, whether or not the mention carries a link. In practice that shifts what you ask for. When a client writes a LinkedIn post about a project, the name in plain text is worth something even without a URL. When you speak at a meetup and the organiser writes it up, that counts. Neither of those would have registered as an SEO activity five years ago.

Treating ChatGPT, Google AI Overviews and Perplexity as a single channel called "AI search" is a category error. Analysis of the citation data found only 6.8% of cited domains appeared on three or more platforms. The engines are reading materially different parts of the web. Wikipedia dominates ChatGPT's top sources. Reddit dominates Perplexity's. Google AI Overviews lean on Reddit, YouTube and Quora. A page that gets you cited in one may be invisible to the others, so a single "get cited by AI" project is really three projects with different source mixes. Profound's separate analysis of roughly 730,000 ChatGPT conversations from October to December 2025 found the citation distribution had a Gini coefficient of 0.8, with the top 10 domains accounting for only 12% of all citations. Their summary was: "No single source dominates. But the distribution is highly unequal... Everyone has a chance; few win big." That is genuinely good news for a small studio. There is no gatekeeper domain you must appear on. The long tail is enormous, which means a niche page that answers a specific question well can be pulled into an answer without competing against a national publisher for the slot.

The same technology that occasionally sends you a pre-qualified enquiry is reducing the clicks your existing rankings earn. Ahrefs analysed 300,000 keywords in 2025 and found the top-ranking page had a 34.5% lower average click-through rate when an AI Overview was present, compared with similar informational keywords without one. A later Ahrefs figure published in December 2025 put the reduction at 58% for position one. The two numbers are not contradictory. They are different samples at different points in a rollout that was still expanding, and the direction of travel between them is the point. So the net position for a service business is not "AI search is a new channel that adds traffic". It is closer to "informational search traffic is shrinking, and a small, high-intent replacement stream is appearing alongside it". The replacement is nowhere near the size of the loss at current volumes. This is why we would not advise anyone to fund AI search work out of enthusiasm. Fund it because the same work, publishing specific and genuinely useful answers to questions your buyers ask, is the thing that survives whichever way retrieval goes next.

Australian use of AI assistants is now mainstream, which is what makes the referral path plausible rather than exotic. Roy Morgan found that 58% of Australians aged 14 and over used AI tools in an average four weeks during the March quarter of 2026, with 10.5 million Australian ChatGPT users, the largest single tool. Broad adoption is not the same as broad use for supplier selection. Most of that usage is drafting, summarising, homework and curiosity. The share of Australians who have asked an assistant to recommend a software developer or an automation consultant is unknown, and we could not find a credible figure for it. What the adoption number does tell you is that the behaviour is available to almost everyone you sell to. A Melbourne agency principal who wants a quoting workflow built has an assistant open on their laptop already. Whether they use it that way this year or next is a question of habit, not access. That is the reason to do a small amount of work now rather than none. The cost of being findable in AI answers is mostly the cost of writing clearly, which you should be doing regardless.

The published statistics on AI search are unusually unreliable, and you should treat almost any confident percentage you read about it with suspicion. While researching this guide we tried to trace the most widely repeated figures back to a primary source and could not. These are the specific claims we could not stand behind, and did not use: - "AI referral traffic converts at 4.4x to 23x organic." Attributed to Semrush, Seer Interactive and Ahrefs across many blogs. We found no Semrush or Ahrefs publication stating either multiple, and the arithmetic does not reconcile with the Seer data it supposedly derives from, since 15.9 divided by 1.76 is about 9. - "AI traffic grew 16x" and "AI referrals grew 527%." Both circulate widely with no traceable primary study behind either. - Listicle share of citations, quoted variously as 43.8%, 63%, 40% and 25%. Every version we chased led to an aggregator post rather than a study. The Profound research these are attributed to publishes domain-level data, not content-format breakdowns. - "Self-contained passages get 2.3x more citations." No traceable study. The underlying mechanism is real, which we explain below, but the number is not. The pattern is blogs citing blogs until a number acquires the appearance of a source. If you are being pitched AI search services, ask for the primary study behind every figure in the deck.

Write pages that answer one specific question completely, then get your name onto pages you do not control. That is the whole practical programme, and both halves are supported by the evidence above rather than by vendor enthusiasm. The first half rests on how retrieval works. These systems do not read your page as a document; they split it into passages and retrieve the passages that match the question. A section whose opening sentences answer the question on their own can be lifted into an answer intact. A section that opens with setup, context or a promise of what is coming next gets skipped, because the retrieved chunk says nothing useful without the paragraphs around it. We are not going to attach a multiplier to that, because the ones in circulation are invented, but the mechanism follows directly from how the retrieval step is built. The second half rests on Profound's domain data and the Ahrefs correlation study. Ask satisfied clients to write publicly about the work in their own words on their own accounts. Answer questions in the communities your buyers actually use rather than the ones marketers talk about. Keep your listings in industry directories and knowledge bases accurate and consistent, since practitioners agree this matters even though we found no controlled study measuring it. Then measure it honestly. Add a "how did you hear about us" field to your enquiry form and read the answers. That is how we know about the Melbourne agent, and it cost nothing.

Should I be optimising my website for ChatGPT? Only in the sense that clear, specific, self-contained writing helps. The most-cited domains in Profound's 680 million citation dataset are Wikipedia, Reddit, YouTube, Quora, Forbes, G2 and Yelp, not vendor websites. Your own site is rarely the source an assistant quotes when recommending a business. Getting mentioned accurately on other people's pages does more than any change to your homepage. Is AI search traffic worth chasing yet? For most Australian service businesses, it is worth a small deliberate effort and nothing more. In the Seer Interactive dataset, AI assistants sent about 11,000 sessions against about 14,000,000 Google organic sessions on the same site, roughly 0.07% of organic volume. The visitors convert well but there are very few of them. Do backlinks still matter for AI visibility? They matter less than mentions, on the best available evidence. Ahrefs found branded web mentions correlated with Google AI Overview visibility at 0.664 across 75,000 brands, against 0.218 for backlinks. Ahrefs note, correctly, that correlation is not causation. Treat mentions as a signal worth building rather than a proven lever. Will AI Overviews reduce my existing Google traffic? Probably, if you rank for informational queries. Ahrefs measured a 34.5% lower average click-through rate for the top-ranking page when an AI Overview was present, across 300,000 keywords in 2025, with a later December 2025 figure of 58% for position one. Commercial and navigational queries appear less affected than informational ones. How do I know whether an AI assistant sent me an enquiry? Ask on the form. Referral data from AI assistants is unreliable, because many people read an answer, then type your name into a browser, which lands in your analytics as direct or branded search. A free-text "how did you hear about us" field is the only method that reliably catches it, and it is how we caught ours.