What changed in search
A Google search for a question now often returns an AI Overview above the links, and AI Mode gives a full conversational answer with a handful of cited sources. Bing Copilot does the same from the Bing index. Perplexity does it from its own. The buyer reads the answer and clicks one or two sources, if any.
Those sources are not simply the top ten results. Google's Search Central documentation describes a technique it calls query fan-out: the engine issues several related searches on sub-topics, collects pages that answer each one, and assembles the response. Industry analysis of millions of AI Overview citations in 2026 found only around a third of cited pages ranked in the top ten for the typed query. The rest earned their place by answering a sub-question well.
That is the opening for a small site. You will not outrank a national brand for "shopify migration agency". You can be the page that answers "how long does a Magento to Shopify migration take" clearly, and get cited when the engine asks itself that question on the way to its answer.
What Google says is needed, and what is not
Google is unusually clear here. Its AI features documentation, updated in December 2025, says there are no additional requirements to appear in AI Overviews or AI Mode. The checklist it gives is ordinary: crawling allowed in robots.txt and by your CDN or host; content reachable through internal links; good page experience; important content in text rather than only in images, video or scripts; structured data that matches the visible page; Business Profile and Merchant Center kept current.
It also says, in writing, that you do not need to create new machine-readable files, AI text files or markup for these features. The widely promoted AI text file is not read by any major engine according to server-log studies. FAQ rich results were removed for all sites in May 2026, and HowTo rich results in 2023. A controlled study in 2026 found that adding structured data did not raise AI citations.
So if a proposal leans on special files, special schema or an "AI-ready" plug-in, it is selling you the parts that do not matter.
What we do
Classic SEO still underpins it
- Crawl access checked at the server and the CDN, including the search-index crawlers of Bing, Perplexity, ChatGPT and Claude that a "block AI bots" switch quietly removes.
- One canonical per page, a single host, a clean sitemap, Search Console and Bing Webmaster Tools both verified, IndexNow set up for Bing and Copilot.
- Core Web Vitals within Google's thresholds: largest contentful paint under 2.5 seconds, interaction to next paint under 200 milliseconds, layout shift under 0.1, each at the 75th percentile.
- Mobile parity: the content a phone sees is the content that gets indexed.
Answer-first page structure
Each page is rebuilt around one question. The question is the heading. The direct answer sits in the first screen in plain text: what it is, who it is for, what it costs, how long it takes. The detail, proof and process follow. Headings are the sub-questions a buyer would ask next, each followed by a direct answer, which is the structure fan-out rewards. Long "ultimate guides" are split into focused pages.
Fan-out sub-pages
The audit lists the sub-questions the engines generate for your topic and the ones your buyers type: cost, duration, comparison, what breaks, what is included. Each missing question gets one focused page, written with you from your own facts.
Entity consistency
Your business name, description, founder, founding year, address and phone are made identical across the site, the Organization markup, your Business Profile, and every directory or review profile. Engines reconcile what they read against this record, and buyers read inconsistency as risk.
Ecommerce pages readable by AI shopping answers
Product and category pages get plain-text titles, prices, availability, specifications and a one-paragraph description of who the product suits. Product structured data is checked against the visible page and against the Merchant Center feed, so an AI shopping answer quoting your price quotes the right one.
How we read the Search Console generative AI report
Search Console now has a report, under Performance, for AI features in Search. It shows impressions only: how often your pages appeared inside AI Overviews and AI Mode, by page, country, device and date. It does not show the queries, the clicks or the position, and it is not available through the API.
We use it as a visibility counter, not a traffic report. Three questions it answers well: which of your pages Google is willing to pull into AI answers at all; whether the impressions come from your target country or from noise; and which pages appear in AI features but earn no clicks in the ordinary performance report. Those cited-but-not-clicked pages get their first screen rewritten so that there is a reason to open them. We pair the report with the twenty-question runs, which do show the actual answers and sources.
Illustrative record
A US WooCommerce store selling specialist equipment, strong on classic rankings for product names, invisible for the comparison and how-to questions its buyers ask. Illustrative figures showing the shape of a baseline and second measurement, not a client result.
- Baseline: cited in 4 of 80 question-engine cells; AI features impressions concentrated on 2 blog posts; 6 category pages with prices only in images; CDN rule blocking two search-index crawlers; three versions of the business description across the web
+ After the sprint: 9 pages restructured answer-first, 5 cost and comparison pages written, prices and specifications in text on every category page, crawler rule removed, one description everywhere, product data matched to the feed
What it costs
| Work | Includes | Fixed price |
|---|---|---|
| AI visibility checkThe first step | Five questions, four engines, written report of who is cited and whether you are. 2 working days | Free |
| AI Visibility Audit | Twenty-question baseline, Search Console and crawler read, page structure review, fan-out question list, priced plan. 10 working days | $2,500credited against the sprint |
| Implementation sprint | Technical fixes, answer-first restructuring, fan-out sub-pages, entity consistency, product data, second measurement. 4 to 8 weeks | $3,500 to $9,000 |
| AI Visibility WatchOptional, cancel any month | Monthly question runs, generative AI report review, follow-on fixes, a named engineer | $300 to $1,200 a month |
What moves the sprint price: how many pages need restructuring, how many sub-question pages are missing, the size of the catalog if you sell online, and the technical state of the site. The price is written down before work starts. Half at start, half at the second measurement.
How it runs
Free check (2 working days)
Send your site address and five buyer questions. We run them on Google AI, Copilot, Perplexity and ChatGPT and send a written report of the answers, the cited sources and whether you appear.
Audit (10 working days)
Twenty questions agreed and run as a baseline. Search Console, Bing Webmaster Tools and server access reviewed. Every page read for structure and missing facts. A plan with each fix, its expected effect, effort and order, priced as a sprint.
Sprint (4 to 8 weeks)
Technical fixes first, then page restructuring and new pages on a staging copy you can open any time, with a written log every working day. You approve each page before it goes live.
Second measurement
The twenty questions re-run from the same location on the same engines, laid beside the baseline, with the first generative AI report readings. We walk through it with you.
Watch (optional)
Monthly runs, report review and follow-on fixes from a named engineer. Cancel any month.
Who does this work
RiantSoft is an engineering firm, founded in 2009, that has spent seventeen years building, upgrading and maintaining ecommerce stores and business applications for US and UK clients. Answer engine work is page structure, data correctness, crawler access and measurement, which is the work we already do. Our own site was rebuilt on these rules in October 2026, and we are measuring its baseline the same way we would measure yours.
Get a free AI visibility check, or email info@riantsoft.com.
Questions people ask
Is answer engine optimization just SEO with a new name?
On Google, largely yes, and Google says so. Its documentation states that AI Overviews and AI Mode use the same index, the same ranking systems and the same eligibility rules as ordinary search, with no extra requirements. What changes is which pages get picked: the answer is assembled from several related searches, so pages that answer a narrow sub-question directly are chosen over pages that rank for the broad term. The technical foundation is the same SEO you know. The page structure and the question coverage are what most sites lack.
Do we need to rank on page one to appear in an AI Overview?
Less than you would think. Industry analysis of millions of AI Overview citations in 2026 found that only around a third of cited pages ranked in the top ten for the typed query. The rest ranked lower, or for a related sub-question the engine generated on its own. That is good news for a small site: you cannot outrank a national brand for the head term, but you can own the specific question the engine asks on the way to its answer.
Should we add FAQ schema or an AI text file?
No. Google removed FAQ rich results for all sites in May 2026, and its AI features documentation says no special files, AI text files or markup are needed. Keep the FAQ content, because question headings followed by direct answers are exactly what the engines reward; drop the expectation that the markup earns anything. We implement Organization, Breadcrumb and Article markup because they are supported and keep your entity consistent, and nothing beyond that.
How do you measure AI Overview visibility?
Two ways. We run your twenty questions logged out from a US location and record whether an AI answer appears, whether you are cited, and which page. And we read the Search Console generative AI report, which shows how often your pages were shown inside AI features by page, country and date. That report shows impressions only, no queries or clicks, so we treat it as a visibility counter and pair it with the question runs to see what is actually being answered.
Will being cited bring traffic or just exposure?
Both happen, and the split varies by question. A cited page whose first screen states a price, a timeline and a next step earns the click because the reader wants the detail. A page that is cited for a definition often does not, because the answer was complete. We identify the cited-but-not-clicked pages from the two reports and rewrite the first screen so there is a reason to open it. Exposure in a shortlist still counts: the buyer who saw your name three times will search for it.
Does this work for an ecommerce store?
Yes, and it is one of the least crowded areas. Shopping answers in Google AI Mode, Copilot and Perplexity read product and category pages, so those pages need plain-text names, prices, availability, specifications and a one-paragraph description that says who the product is for. Merchant Center feeds must agree with the page. We also run product-level buyer questions in the measurement, not only 'best store for' questions. If you are moving platforms, the redirect and structure work belongs in the migration.
What happens after the sprint?
You get the second measurement, a written record of every change, and the choice of a Watch plan. Watch re-runs the twenty questions monthly, reviews the Search Console generative AI report, and makes the follow-on fixes, with a named engineer who already knows your site. It is $300 to $1,200 a month depending on page count and can be cancelled any month. Nothing is locked in, and everything lives in your own accounts.