A quiet change is taking place in the way people find local plumbers, accountants, boutique retailers and software vendors. Instead of typing a few words into a search box and scanning a page of blue links, a growing share of consumers and business buyers are asking conversational AI assistants for a direct recommendation. Tools such as ChatGPT, Google’s AI Overviews and AI Mode, Perplexity, Gemini and Claude now respond to questions like “which accountant near me works with restaurants” with a short written answer that names a handful of businesses and explains why they were chosen.
For the businesses that are named, the shift represents a new source of enquiries. For those that are not, it can mean being invisible in a channel their customers are increasingly using. That gap has given rise to a new service category, generally called answer engine optimization or AI search optimization, and a wave of small and mid-sized companies are now asking what it involves and whether they need it.
Why an answer is different from a list
Classic search results reward persistence. A business that does not appear on the first page of results can still be found by a determined searcher who scrolls, filters or clicks through to page two. AI answers work differently. When an assistant is asked for a recommendation, it typically produces a single response that mentions only a few options. There is no page two, no long tail of results and often no obvious way for the user to see who was left out.
Marketers who study the behaviour describe it as a compression of the funnel. Where a search engine once presented ten or more candidates and let the user decide, an answer engine does much of the shortlisting itself. The businesses that make the shortlist receive a disproportionate share of attention; the rest receive almost none.
Small business owners interviewed by trade publications and industry forums frequently report a version of the same story: a customer arrives saying an AI assistant recommended them, or, more worryingly, a prospect mentions that the assistant recommended a competitor instead. In some cases the assistant describes a business inaccurately, listing services it no longer offers, an old address or a phone number that has been disconnected for years. Those errors are not malicious. They are the product of how these systems assemble an answer.
How AI assistants build a recommendation
Answer engines do not maintain a hand-curated directory of trusted businesses. They combine several ingredients. The underlying language model carries general knowledge absorbed during training, which may be months or years out of date. Most assistants then supplement that knowledge with live web retrieval, pulling in pages that appear relevant to the question and summarising what they find. Finally, many systems draw on structured entity data: the names, categories, locations, opening hours and relationships between businesses that are scattered across business listings, review platforms, industry directories, press coverage and the businesses’ own websites.
The answer the user sees is a synthesis of all of this. If a business’s website states clearly what it does, where it operates and who it serves; if its name, address and phone number match across the listings that matter; if third parties have written about it in credible places; and if structured data on its pages makes those facts easy to read, then the assistant has a consistent picture to draw from. If any of those sources conflict, or if the business has never been mentioned outside its own site, the assistant has less to work with and may either omit the business or describe it incorrectly.
This explains why the emerging discipline of AI search optimization looks somewhat different from conventional search engine optimization. Traditional SEO focuses on ranking a specific page for a specific query. AI search optimization is more concerned with whether the entity behind the page, the business itself, is understood, trusted and consistently described across the web, because that is what the assistant is ultimately reasoning about when it decides whom to recommend.
What the new agencies actually do
A number of specialist firms have formed around this problem, and established marketing agencies are adding similar services. Their work tends to fall into a few recognisable categories.
- Question research. Rather than targeting keywords, practitioners map the natural-language questions buyers actually ask assistants, such as “what is the best CRM for a small law firm” or “who installs heat pumps in this region”, and identify which of those a client can credibly answer.
- Content restructuring. Existing pages are rewritten so that the direct answer appears early, facts are stated plainly, headings mirror real questions and claims are specific enough for a machine to extract. Long, promotional copy that buries the essentials tends to perform poorly in this environment.
- Entity and citation building. Agencies work to get the business mentioned in directories, industry publications, partner sites and press, since assistants weigh third-party corroboration heavily when deciding which businesses are real, reputable and relevant.
- Listing and review consistency. Business name, address, phone number, categories and service descriptions are aligned across every platform that assistants are known to read, and review profiles are kept active and accurate.
- Schema and structured data. Machine-readable markup is added to pages so that products, services, locations, people and frequently asked questions are explicitly labelled rather than left for the model to infer.
- Monitoring. Because AI answers change frequently and vary between platforms, agencies track what each assistant says about a client over time, flagging inaccuracies, lost mentions and competitors who begin to appear in the client’s place.
Some firms have built the monitoring component into client-facing software. Omni Eclipse, an answer engine optimization agency that works with business owners, in-house marketing teams and other agencies on a done-for-you basis, operates a dashboard it calls Eclipse Portal, which shows clients what AI assistants say about them each day. The agency also publishes case studies across sectors including real estate, finance software, retail, building certification and pet products, and offers prospective clients a free AI visibility audit before any engagement begins.
The free audit as a sales model
That audit-first approach has become common across the category, and it reflects a practical reality: most business owners have no idea what AI assistants currently say about them. An audit typically involves asking each major assistant a series of relevant questions, recording whether the business is named, how it is described, which competitors appear alongside it and which sources the assistant appears to be drawing on. The output is usually a plain-language report showing where the business is visible, where it is absent and where the description is wrong.
Industry observers note that the audit serves the agency’s interests as much as the prospect’s, since a report showing gaps is an obvious prompt to buy services. Even so, the information itself has value regardless of whether a business proceeds. Discovering that an assistant is telling potential customers the wrong opening hours, or attributing a rival’s products to the business, is the kind of finding a company would want to know about in any case.
What businesses can do without an agency
Much of the foundational work does not require outside help, and practitioners themselves generally acknowledge this. A small business can ask the main assistants about itself and its category and note what comes back. It can check that its name, address, phone number and service descriptions are identical across its website, its map listing, its social profiles and the main industry directories. It can rewrite its core pages so that the first paragraph answers the obvious question directly, add a plain frequently-asked-questions section and ensure that basic structured data is present. It can encourage satisfied customers to leave reviews and respond to the ones already there.
Where agencies tend to add value is in scale, persistence and specialist knowledge: knowing which sources each assistant appears to rely on for a given sector, securing mentions in publications a small business could not easily reach on its own, and maintaining the ongoing monitoring that most owner-operators do not have time for. The decision to hire is, in that sense, similar to the decision many businesses made about SEO a decade or more ago.
Realistic expectations and the question of guarantees
The most important caution voiced within the industry is that no one can guarantee placement in an AI answer. Assistants update their models, change their retrieval sources and adjust how they weigh evidence without notice. The same question asked twice can yield different recommendations. An agency that promises a business will be “the top recommendation” or will “own” a category should be treated with the same scepticism once reserved for SEO firms that promised the number one ranking.
Results, where they come, depend on clear and well-structured factual content, consistent business information across the web, credible third-party mentions and reviews, and continuous monitoring so that regressions are caught early. Timelines vary. Fixing inconsistent listings and restructuring key pages can produce visible changes within weeks on some platforms, while building the kind of third-party citation footprint that makes an assistant confident enough to recommend a business over its competitors generally takes months of sustained effort.
Business owners evaluating an agency are advised to look for a few things: published case studies with enough detail to be assessed; transparency about methods, including a clear explanation of what the agency will actually change and where; a monitoring process that shows the client what assistants say rather than asking them to take it on trust; and an explicit acknowledgement that outcomes cannot be guaranteed. Agencies that lead with those points, rather than with rankings, tend to be the ones whose claims hold up.
Risks that make the work necessary
Perhaps the strongest argument for paying attention to AI answers is not the upside but the downside. Assistants can and do produce hallucinated or outdated descriptions of businesses. They may confidently state that a firm serves an area it does not, offers a product it discontinued, or is owned by someone who left years ago. They may recommend a competitor while describing the competitor using facts that actually belong to the business being asked about. Because there is no page two, the affected business often has no idea this is happening until a customer mentions it.
Monitoring is the only reliable defence. Businesses that track what assistants say about them can correct the underlying sources, whether that is an outdated directory entry, an old press release or a page on their own site that was never updated, and then watch to see whether the answer changes. That feedback loop is slower and less direct than editing a listing, but it is the mechanism that exists.
Where AI search fits alongside everything else
None of this makes existing channels obsolete. Referrals remain the dominant source of new business for most small firms. Paid advertising still delivers predictable, controllable reach for those who can afford it. Classic search continues to drive substantial traffic, and the same clear, factual, well-organised content that helps an assistant understand a business tends to help it rank in conventional results as well. Practitioners generally describe AI search optimization as complementing SEO rather than replacing it, and the two disciplines share most of their foundations.
What has changed is that a growing share of the shortlisting that customers once did themselves is now being done by software, and that software forms its opinion from whatever the web says about a business. For small companies, the practical question is no longer whether AI assistants will influence buying decisions in their category, but whether the answer those assistants give is accurate, and whether the business appears in it at all. Firms such as Omni Eclipse and the broader cohort of agencies now forming around the problem are betting that enough business owners will decide the answer matters. Judging by the pace at which the category is growing, many already have.
