Generative engine optimization: get your firm named when buyers ask an AI who to hire

What generative engine optimization is

When a buyer asks ChatGPT "who should I hire to move my store to Shopify" or asks Perplexity "which accessibility audit firm should I use for a WooCommerce site", the assistant writes a short list with a few names and reasons, and points to the sources it used. That list is now a shortlist. If your firm is not in it, the buyer did not reject you; the buyer never saw you.

Generative engine optimization is the work of getting into that list, honestly and measurably. It covers the assistants that generate answers (ChatGPT, Perplexity, Gemini, Claude) and the Google surfaces that do the same thing inside search (AI Overviews and AI Mode). The page for the Google side specifically is answer engine optimization. This page covers the full job.

What the evidence says gets cited

Industry studies of millions of AI citations, 2025 to 2026, agree on four things.

• Third-party sources beat your own site. When an assistant recommends a vendor it cites that vendor's own website a minority of the time. It leans on directory category pages, review platforms, comparison articles, forums and news. For supplier-selection questions, sector directories alone supply a large share of citations, and the list page is cited far more than any single profile on it.
• Being named matters more than being linked. Across tens of thousands of brands, the count of branded mentions on other sites correlated with AI visibility roughly three times more strongly than backlinks did. Review count correlated with being recommended; the star rating barely did.
• Direct answers near the top get used. Pages that state the answer plainly in the first third of the page, with specific facts and named entities, are cited more than long guides that build up to a point.
• Recent content is favoured. Pages updated within the last few months are cited more often, and Gemini in particular weights recency because it grounds on live search.

One more finding shapes how we work. Google's Search Central documentation states that no special files or markup are required to appear in AI Overviews or AI Mode, and a 2026 controlled study found that adding structured data did not increase citations. So the work is not a technical trick. It is page structure, consistency, presence and measurement.

What we do on your site

• Answer-first pages. Each service or product page is rebuilt so the question it exists to answer is the heading, the direct answer (what, for whom, price range, timeline) sits in the first screen in plain text, and the detail follows. One page, one question.
• The missing pages. The audit finds the questions your buyers ask that nobody on your site answers: "how much does X cost", "how long does X take", "X or Y". Each gets one focused page, written with you.
• Entity consistency. Your legal name, description, founder, founding year, address and phone are made identical across your site, your Organization markup and every profile that describes you. Inconsistency is the thing assistants reconcile against, and the thing buyers read as risk.
• Facts as text. Prices, timelines and specifications are stated as static text, not inside calculators, images or scripts that engines cannot read and routinely misquote.
• Access. We check robots.txt and, more importantly, your CDN or firewall. The common silent failure is a "block AI bots" toggle that also blocks the search-index crawlers of ChatGPT, Perplexity and Claude. Blocking a training crawler is your choice; blocking the search crawlers removes you from citations.
• Merging thin pages. Twenty pages with substance outperform two hundred near-duplicates in classic search and in AI answers, and bulk pages are now covered by spam policy.

What we do off your site

• Directory presence for your category. The directories that assistants cite differ by industry. We identify the ones cited for your questions, complete your profiles with the same facts as your site, and work towards the category list pages, which depend on verified reviews and profile completeness.
• Reviews where engines read them. Review count on the platforms the engines draw from matters more than rating. We set up the request flow so your real customers leave reviews where they count.
• Comparison articles and lists. For each of your twenty questions we show which articles and lists are cited, who publishes them, and what it would take to be included. Where it is a matter of asking, we draft the request. Where it requires original data, see below.
• Original data. The cheapest citable asset a small firm can produce is a real number: a survey of your market, an audit of thirty competitors, a measured result. We help you structure and publish one dataset so that it can be quoted in a sentence.
• Honest forum presence. Reddit and similar communities are among the most cited sources on every engine. We identify the threads where firms like yours are discussed; you answer them as yourself. Nothing planted, nothing paid.

What we refuse to do

• Bought or planted mentions. Google's spam policies now explicitly cover attempts to manipulate AI responses, and inauthentic mentions are treated like bought links. Paid content is also a fraction of a percent of AI citations, so it does not work even before it gets you penalised.
• Pages generated in bulk, city-swap pages, or a second site with the currency changed.
• Special files and markup sold as the fix. Google's documentation says they are not needed; server-log studies show the much-promoted AI text files are not read by any major engine. If a vendor's plan is built on them, the plan is built on nothing.
• A guarantee. We sell measurement and the work the evidence supports.

How we measure

Every engagement runs on the same instrument. Twenty questions, agreed with you, phrased the way your buyers phrase them. Four engines: ChatGPT, Perplexity, Gemini and Google AI Overviews or AI Mode, with Claude added when your buyers use it. Run logged out from a US location so no history shapes the answer. For each question and engine we record whether you are named, whether you are cited, which page of yours is cited, and which other firms and sources appear.

That gives a baseline table of 80 cells. The sprint ends with the same run, and the Watch plan repeats it monthly. You see the movement, cell by cell.

What it costs

What moves the sprint price: the number of pages to restructure, how many new pages the audit calls for, how many directories and lists matter in your category, and whether a dataset is produced. The price is written down before work starts. Half at start, half at the second measurement.

How it runs

1. Free check (2 working days)

   You send your site address and five buyer questions. We run them on four engines and send a short written report: the answers, who is cited, from which sources, and whether you appear.

2. Audit (10 working days)

   We agree the twenty questions with you, run the full baseline, map the sources cited for your category, read your site page by page, and deliver a plan with every fix, its expected effect, its effort and its order, priced as a sprint.

3. Sprint (4 to 8 weeks)

   Pages are restructured and written on a staging copy you can open any time, with a short written log every working day. Off-site work runs in parallel from your own accounts. You approve every change before it is published.

4. Second measurement

   The twenty questions are re-run on the same engines from the same location and laid next to the baseline. We go through it with you cell by cell and agree what to do next.

5. Watch (optional)

   Monthly re-measurement, a one-page report and the fixes that follow from it, with a named engineer who knows your site. Cancel any month.

Why an engineering firm does this

Most of this work is engineering: page structure, data consistency, crawler access, measurement discipline. RiantSoft has built and maintained ecommerce stores and business applications for US and UK clients since 2009, and its first clients came through search. We rebuilt our own site on these principles in October 2026 and measured our own baseline before launch. The results for our own questions will be published here as they come in.

Get a free AI visibility check, or email info@riantsoft.com.

Questions people ask