Estimated reading time: 6 minutes · Last updated:
To be recommended by ChatGPT and other AI search tools you must make your business visible to the machines that assemble answers. Practically, that means five things: let AI crawlers read your site, win normal search ranking, put the answer in the first one or two sentences, secure third‑party mentions the AI trusts, and keep your business name, address and phone consistent across listings (NAP). GOJI published its findings on 16 Oct 2026 and scored 1,459 AI answers across 100 businesses in 10 industries, finding small businesses appeared in about 5% of those answers. These steps improve AI search visibility and are mostly work SEO teams already do.
Key takeaways
- Dataset: GOJI scored 1,459 AI answers about Australian service costs across 100 businesses in 10 industries and found small businesses in about 5% of them.
- Five steps: Five practical steps to win AI recommendations are: allow crawlers, rank in regular search, answer questions in the opening lines, earn mentions on trusted third‑party sites, and unify your NAP.
- Quick check: A simple monthly test runs ten customer questions in ChatGPT, Perplexity, Gemini and Google's AI Mode and records whether you were named and which sources were cited.
- Fast wins: Verifying your site in Bing Webmaster Tools takes about 15 minutes and can unblock other AI tools that draw on non‑Google indexes.
Table of contents
Why AI names the businesses it can read and verify
AI search tools build answers from the web pages they can find, read and cross‑check. They rarely invent recommendations: instead they harvest content from sites, directories, review platforms, industry bodies and news items, and then summarise those sources. That means the limiting factor is not quality alone but machine readability and corroboration.
GOJI’s audit of AI answers shows this effect in numbers: its dataset of 1,459 AI answers covered 100 businesses across 10 industries, and small businesses appeared in about 5% of those answers. In other words, even good local providers can be invisible if they lack crawlable pages, structured data or trusted third‑party citations.
Practical consequence: optimise for the content the AI actually consumes — the HTML it can parse, the structured data it reads and the third‑party pages it can use to verify facts. Treat AI visibility as an extension of search indexability and trust signals, not as a separate mystery.
The five steps that actually win recommendations
1) Let crawlers in: check robots.txt and any security rules so OpenAI’s and other crawlers can read pages. OpenAI lists its crawlers on its crawler documentation page; verifying crawler access is a one‑time fix that prevents accidental blocking.
2) Rank in normal search first: AI layers on top of regular SEO. If your page isn’t indexed and ranking for the customer question, AI tools won’t quote it. Follow Google’s guidance for AI features so pages are eligible to appear in search.
3) Answer up front: write each service page so the response appears in the first one or two plain sentences. AI tools tend to lift the opening lines, so state what you do, where and a price or price range if possible.
4) Earn trusted mentions: AI answers often cite directories, review platforms and media coverage. Search the question in ChatGPT or Perplexity in a private window, note the cited sites and prioritise claiming listings and pitching the publications your customers read.
5) Make your facts machine‑readable: pick one canonical NAP, publish it everywhere and add LocalBusiness schema so machines don’t have to guess.
How to test AI visibility and measure progress
A repeatable test beats luck. Pick ten customer questions — for example “who’s the best [service] in [suburb]?” and “how much does [service] cost in [city]?” — and run them in ChatGPT, Perplexity, Gemini and Google’s AI Mode in a private window. Record whether you were named, who was named instead, and which sources the AI cited.
Do the audit monthly and track trends over three to six months. Single checks vary day to day; GOJI recommends monitoring a trend because third‑party mentions and rankings evolve slowly. When you repeat the exercise, the third column — sources cited — tells you the next sites to target for listings or PR.
If you prefer tool support, agencies and platforms exist that automate rank capture and AI answer tracking. For many small companies, the manual ten‑question check takes a few hours a month and focuses in on the exact gaps the AI is using to exclude you.
AEO versus SEO, timelines and what usually changes
Answer engine optimisation (AEO) builds on search optimisation rather than replacing it. SEO gets pages indexed and eligible; AEO ensures the AI can quote you, trust you and finds you named on corroborating sites. In practice the overlap is large: technical SEO and clear answers are the same work described with a different emphasis.
Timescales are straightforward: crawler and site fixes can show results in weeks, while third‑party mentions, reviews and rankings take months. Track progress monthly and judge the trend over a quarter or two — three to six months typically separates a baseline check from measurable capture of AI answers.
A small content change — putting the answer in the opening lines and publishing prices or ranges — can produce visible movement quickly, but durable AI recommendations still depend on sustained third‑party validation.
How this can play out
The case for
- Firms that fix crawler access and publish clear answers should see gains within weeks because AI tools lift opening lines and indexed pages.
- Earning a small number of high‑quality third‑party mentions can disproportionately increase AI citation because tools prefer corroborated sources.
The case against
- If directories and review platforms remain dominated by national players, small local firms may take months to surface despite clean site changes.
- Blocking crawlers or inconsistent NAP across listings will prevent even well‑ranked pages from being quoted by AI, negating technical and content work.
What to be careful about
- Accidentally blocking OpenAI or other crawlers via robots.txt or security settings, which prevents machines from seeing your pages.
- Relying solely on short content fixes without pursuing third‑party mentions, so AI continues to cite competitors.
- Inconsistent business name, address or phone across sites reducing machine trust and lowering the chance of citation.
The bottom line
AI recommendations favour businesses that are easy for machines to read and verify. The five steps — open crawler access, solid ranking in regular search, concise opening answers, trusted third‑party mentions and consistent NAP with LocalBusiness schema — are practical priorities. Run the ten‑question check monthly, fix crawler or schema issues first, then invest in listings and reviews. GOJI’s dataset (1,459 AI answers across 100 businesses in 10 industries) shows the gap for small providers today; closing it takes ordinary SEO and a focused listings programme rather than a full rebuild.
What to watch
- Run the ten‑question AI visibility check in your next monthly audit; no date has been set.
- Verify your site in Bing Webmaster Tools and confirm crawler access; this verification takes about 15 minutes.
- Track the trend of named results over three to six months after you publish customer‑answering content; no date has been set.
Frequently asked questions
How do AI tools decide which businesses to name?
AI tools aggregate web sources they can find and verify; GOJI’s analysis of 1,459 AI answers across 100 businesses in 10 industries showed that machine‑readability and third‑party corroboration drive citations, and small businesses appeared in about 5% of those answers.
How quickly will changes show up in AI answers?
Crawler and site fixes can appear within weeks, while third‑party mentions and improved rankings generally take months; expect to judge progress over a quarter or two, roughly three to six months.
Do I need a new website to be recommended?
Usually not. Most businesses only need edits that allow crawling, clearer opening paragraphs (one or two plain sentences answering the customer question) and structured data; a rebuild is needed only if the site cannot be crawled or edited.
Related reading
