In this article (9 sections)
- What it means to be cited in an AI answer
- How AI assistants choose which businesses to cite
- Check that AI search crawlers can reach your website
- Write pages that answer the buyer's follow-up questions
- Why AI answers repeat what other websites say about your business
- Keep business details consistent everywhere AI assistants look
- AI citation tactics that waste time and money
- How to measure whether AI answers are citing your business
- Where to start with AI citations
More buyers now ask an AI assistant for a shortlist before they visit a single website. Someone planning a kitchen renovation asks ChatGPT which builders near them handle heritage homes. A finance manager asks Google's AI Mode to compare bookkeeping firms. The businesses named in that answer get considered. The rest are never seen.
Most of the advice about fixing that is sold under a new label, GEO or AEO, and a lot of it is guesswork. We'll stick to what the platforms themselves publish about how they choose sources, what you can actually change as an Australian business, and which of the popular tactics you can safely ignore.
What it means to be cited in an AI answer
"Cited" gets used loosely, and it covers three different outcomes. They're worth separating, because you win each one differently and measure it differently.
| Outcome | What the reader sees | What it is worth |
|---|---|---|
| A mention | Your business named in the text of the answer, with no link | Brand exposure at the moment of choice. The reader has to search for you next. |
| A linked citation | A source link or card pointing at one of your pages | A possible visit, and a sign the assistant trusted your page for a fact |
| A recommendation | Your business listed as an option for the reader's specific need | The closest thing to a referral, and the hardest to earn |
You can be linked as a source for a general fact and never be recommended, or recommended without your own site being linked at all. Recommendations usually lean on what other websites say about you, which is why the work below doesn't stop at your own pages.
How AI assistants choose which businesses to cite
An AI assistant answering a question about the world right now isn't recalling a list of businesses from memory. It searches, reads what comes back, and writes an answer grounded in those pages. Google describes its own version of this plainly in its guide to generative AI features: AI Overviews and AI Mode retrieve pages from the Search index using Google's core ranking systems, then build the response from the information on those pages.
Two consequences follow if you want to be cited.
- If you can't be found in the underlying search, you can't be cited. Google says a page must be indexed and eligible to show with a snippet in ordinary Search before it can appear as a supporting link in its AI features. There's no separate door.
- The assistant runs searches you never see. Google calls this query fan-out: the model issues a set of related searches on subtopics of the original question. A question about choosing a builder for a heritage home might fan out into searches about council approvals, typical costs and what to ask on a site visit. The pages that rank for those follow-up searches are the candidates for citation.
ChatGPT, Claude and Perplexity work on the same principle with their own search systems and their own crawlers. That makes your first practical question a very unglamorous one: can those crawlers get to your site at all?
Check that AI search crawlers can reach your website
Every major assistant lets you control access for its search results separately from use of your content for training. You set the two independently in robots.txt, and mixing them up is one of the most common reasons a business never appears.
| Platform | Feeds search answers | Collects training data | What blocking the search crawler does |
|---|---|---|---|
| Google (AI Overviews, AI Mode) | Googlebot, the same crawler as Search | Google-Extended controls use for Gemini training | Blocking Googlebot removes you from Search entirely. Google says Google-Extended does not affect inclusion or ranking in Search. |
| ChatGPT | OAI-SearchBot | GPTBot | OpenAI says sites that opt out of OAI-SearchBot "will not be shown in ChatGPT search answers". |
| Claude | Claude-SearchBot | ClaudeBot | Anthropic says it stops your content being indexed for search, which may reduce your visibility in Claude's answers. |
| Perplexity | PerplexityBot | Perplexity says PerplexityBot is not used for model training | Perplexity recommends allowing PerplexityBot if you want to appear in its search results. |
Each platform also runs a user-triggered fetcher (ChatGPT-User, Claude-User, Perplexity-User) that visits a page when a person asks the assistant to read it. Block those and the assistant can't open your page even when a customer pastes the link in.
Four checks cover most of your risk:
- Read your
robots.txtat yourdomain.com.au/robots.txt. Look for any group that names the search crawlers above with aDisallow: /. - Watch for the specific-group trap. Google's robots.txt documentation is explicit that a crawler obeys only the most specific group that names it, and that named groups and the
*group are not combined. A group written for one bot quietly overrides your general rules for that bot. - Check your firewall and CDN, not just the file. Security plugins and CDN bot settings can block a crawler before it ever reads robots.txt. On Cloudflare, the Block AI bots setting targets training crawlers and excludes bots used for both training and search, but custom firewall rules and third-party security tools aren't always so careful.
- Confirm the Search Console setting. Google now has a Search generative AI control under Settings in Search Console. Inclusion is the default for every property, but it's worth confirming nobody switched yours off.
Blocking training crawlers is a legitimate business decision, and it's separate from being cited. You can refuse GPTBot and ClaudeBot and still let OAI-SearchBot and Claude-SearchBot in. Our own robots.txt names every one of these crawlers individually, so our policy is explicit rather than inherited from a catch-all rule by accident.
Write pages that answer the buyer's follow-up questions
Once the crawlers can reach you, the question becomes whether your pages are worth citing. Google's guide is direct about what that means. It contrasts commodity content, the "7 tips" article built from common knowledge that anyone or any AI model could produce, with non-commodity content that carries first-hand experience and a point of view. Its advice amounts to this: don't recycle what's already online, because an AI answer doesn't need you to repeat it.
If you run a service business, the gap is rarely your headline service. Most sites describe what they do. Very few answer what a buyer asks next, and those follow-up questions are exactly where fan-out searches go looking.
A practical way to find them:
- List the questions your sales team hears every week. Pricing factors, suitability, lead times, what's included, what goes wrong. These are the real fan-out queries.
- Pull the long questions from Search Console. Filter queries for words like how, what, which, cost and versus. They show the phrasing buyers use.
- Ask the assistants yourself. Put your core buying question to ChatGPT, Perplexity and Google's AI Mode and note which sources they cite, and what those pages explain that yours don't.
- Map each question to one page. A cost question usually belongs on the service page. A comparison or a deeper explanation may earn its own article.
Then answer those questions with the material only your business has. The details that make a page worth citing are the ones your competitors can't copy:
- Who your service suits, and who it doesn't. An honest "this is not for you if" line is rare, useful and very quotable.
- What drives the cost, even if you never publish a price. The factors, and how they move the quote.
- Your actual process, step by step, including the parts customers find surprising.
- Specifics tied to a place: the councils you deal with, the conditions in your area, the regulations your industry works under in your state.
- A named person who stands behind it, with the experience that makes them worth listening to.
Resist the obvious shortcut. Google warns that creating separate pages for every variation of a question, including fan-out queries, mainly to influence rankings or AI answers breaches its scaled content abuse policy. One strong page that covers a topic thoroughly beats twenty thin ones that each cover a slice of it.
Ready to grow faster?
Get a free strategy session with the team behind a 5.0 rating from 120+ reviews.
Get Your Free ProposalWhy AI answers repeat what other websites say about your business
When someone asks an assistant which business to choose, the answer rarely rests on that business's own website alone. It draws on reviews, directories, industry sites, local media and forum discussions, and it tends to name the businesses those sources agree about.
The best public evidence for this is correlational, and you should read it that way. Ahrefs studied 75,000 brands in Google's AI Overviews and found that branded web mentions correlated with AI Overview visibility at 0.664, well ahead of branded anchor text (0.527) and the raw number of backlinks (0.218). The authors stress that correlation is not causation, and the study only included sites above a set authority threshold. The direction is still clear: being talked about, by name, across the web tracks with being named in answers far more closely than link counts do.
If you're an Australian business, the places that genuinely describe you are usually close to hand:
- Your Google Business Profile and its reviews. Google's guide names Business Profiles as one of the ways local businesses become visible in its AI responses.
- Industry associations and licensing bodies, whose member directories say what you do and where.
- Suppliers and manufacturers that list you as a stockist, installer or accredited partner.
- Local and trade media, when you have a real story or a view worth quoting.
- Community discussion, where customers recommend you unprompted.
Most of these are the same relationships that produce good links, which is why the ways Australian businesses earn links without buying them double as a mention strategy. What doesn't work is manufacturing the chatter. Google's guide lists seeking inauthentic mentions among the tactics that are less helpful than they look, and its spam systems sit underneath its AI features in the same way they sit under ordinary results.
Keep business details consistent everywhere AI assistants look
An assistant that finds two addresses, three phone numbers and a service list that disagrees with your own website has a reason to hedge, or to recommend someone whose details line up. Consistency isn't a ranking trick. It removes doubt.
- Name, address and phone identical on your website, Google Business Profile and the directories that matter in your industry.
- Services described the same way on your site and your profile, using the words your customers use.
- Service areas stated plainly in page text, not only in a map embed or an image.
- Hours, licences and accreditations kept current, because stale details are the ones assistants repeat.
If you serve nearby customers, this overlaps almost entirely with the groundwork of local SEO for businesses competing in the map results, and the same fixes serve both.
Structured data helps machines read your details, but keep it in proportion. Google says there is no special schema markup for its AI features and that structured data is not required to appear in them. It's still worth doing for rich results, and it must describe what's visible on your page. Markup that claims something the page doesn't show is a liability, not a shortcut.
AI citation tactics that waste time and money
Google's guide includes a short myth-busting section, and it's worth reading before you sign any proposal sold as AI search optimisation. The tactics below are widely sold. The evidence behind them is thin, or Google has said outright that its systems ignore them.
| Tactic | What the evidence says |
|---|---|
| Publishing an llms.txt file to get into Google's AI answers | Google says Search ignores llms.txt and similar AI text files. It neither helps nor harms. Other tools may read one, which is why we keep a hand-written file on our own site, but it isn't a Google lever. |
| "Chunking" pages into short fragments for AI | Google says its systems understand pages covering several topics and there is no requirement to break content up. Write for the reader. |
| Rewriting every page in a special style for AI | Google says you do not need to write differently for generative AI search, and that its systems understand synonyms without every keyword variation. |
| Hundreds of pages targeting every fan-out question | Named by Google as scaled content abuse when the aim is to manipulate rankings or AI answers. |
| Paying for mentions, listicles and fake reviews | Google names seeking inauthentic mentions as a tactic to ignore, and its spam systems sit underneath its AI features. |
| Tools that claim to measure an "AI ranking score" | Google says no third-party tool has access to its internal ranking or AI systems. Treat any score as the vendor's estimate. |
None of this means GEO is a fiction. It means the part of it that's useful to you is SEO done thoroughly, with extra attention to crawler access, follow-up questions and off-site reputation. Google's own guide says as much: from its point of view, optimising for AI search is still SEO. The practical upshot for you as a buyer is that an SEO campaign built on sound technical and content foundations already carries most of the load.
How to measure whether AI answers are citing your business
There's no single number for AI visibility, and anyone quoting you one is estimating. Four measures, used together, give an honest picture.
- Search Console's generative AI performance report. Google rolled it out to all websites worldwide by 31 August 2026. It shows how often links to your site appeared in AI Overviews and AI Mode, by page, country, date and device. It reports impressions only, with no clicks or positions, and a site needs enough appearances before data shows.
- A fixed prompt sample, checked on a schedule you can keep. Write down twenty to fifty real buying questions, run them in each assistant, and record who is named and which pages are linked. Answers vary between runs, so look for a pattern across the sample rather than reacting to a single response.
- Referral sessions in your analytics. Visits referred from chatgpt.com, perplexity.ai, claude.ai, gemini.google.com and copilot.microsoft.com show up as referral traffic. The volumes are usually small. Look at which pages they land on and whether those visitors enquire.
- Ask new customers where they found you. Add AI assistant as an option on your enquiry form or intake call. Some buyers who first meet you in an AI answer search your name afterwards, so their visit looks like branded search and the assistant gets no credit anywhere else.
Judge the work on the last two. A mention isn't a visit, and a visit isn't a lead. The point of being cited is the enquiry that follows it.
Where to start with AI citations
If you only have the time or budget for part of this, work through it in order. Each step depends on the one before it.
| Step | The work | Why it comes here |
|---|---|---|
| 1. Access | Allow the AI search crawlers, check firewall and CDN rules, confirm the Search Console setting | Nothing else counts if the assistant can't read your pages |
| 2. Findability | Fix indexing and technical problems on the pages that earn money | AI answers are built from search results |
| 3. Substance | Answer buyers' follow-up questions with detail only you have | Commodity content gives an assistant no reason to cite you |
| 4. Consistency | Align your business details across the site, profile and directories | Conflicting details make assistants hedge |
| 5. Reputation | Earn genuine mentions from associations, suppliers, media and customers | Recommendations lean on what others say about you |
| 6. Measurement | Search Console report, prompt sample, referrals, customer answers | Shows which of the above is actually moving |
Step one is where the avoidable misses hide. Steps three and five are where most of the ongoing work sits. If you'd like a second opinion on where your business stands, this is the work behind how we help businesses get named in AI answers, and it sits alongside everything else we do as an Australian digital marketing agency. Whoever does it, start with the crawler check. It costs you nothing and it's the one problem that makes every other effort invisible.
Frequently asked questions
Is GEO different from SEO?
Mostly in emphasis. Google says that from its point of view, optimising for its generative AI features is still SEO, because those features draw on its core ranking systems. The extra attention GEO adds is on crawler access for other assistants, answering buyers' follow-up questions, and how other websites describe your business.
Can you pay to be cited in ChatGPT or Google's AI Overviews?
Advertising is sold separately and labelled as advertising. The sources an answer cites come from the assistant's own search results, which is why crawler access, useful pages and a genuine reputation are the route in, not a placement fee.
Does an llms.txt file help a business get cited?
Not in Google. Google says its Search, including AI Overviews and AI Mode, ignores llms.txt and similar AI text files, so one neither helps nor harms there. Some other tools may read the file, so keeping a short, accurate one is harmless, but it isn't a substitute for the work in this article.
Why does ChatGPT name our competitors but not us?
Check three things in order. First, whether OAI-SearchBot can reach your site, through robots.txt and any firewall or CDN rules. Second, whether your pages rank in ordinary search for the follow-up questions buyers ask. Third, how often other sites, directories and reviews describe your business by name compared with the competitors being named.
Does schema markup get a business cited in AI answers?
Google says structured data is not required for its AI features and there is no special markup for them. It's still worth using for rich results and to state your business details clearly, as long as the markup matches what is visible on the page.
Should we block AI training bots?
That's a business decision, and it's separate from being cited. OpenAI, Anthropic and Google each let you refuse training use while still allowing the crawlers that feed search answers, so blocking GPTBot, ClaudeBot or Google-Extended doesn't by itself remove you from their search results.
The short version
AI assistants build answers from search results, so being cited starts with being reachable and findable. After that it comes down to substance and reputation: pages that answer buyers' real follow-up questions with detail only you have, and other sites that describe your business consistently. Skip the hacks Google says it ignores, and judge the work on enquiries, not mentions.
Ready to grow faster?
Get a free strategy session with the team behind a 5.0 rating from 120+ reviews.
Sources
- Google Search Central: Optimising your website for generative AI features on Google Search
- Google Search Central: AI features and your website
- Search Console Help: Search generative AI control
- Search Console Help: Generative AI performance report (Search)
- Google Search Central: Google's common crawlers, including Google-Extended
- Google Search Central: How Google interprets the robots.txt specification
- OpenAI: Overview of OpenAI crawlers
- Anthropic: Does Anthropic crawl data from the web, and how can site owners block the crawler?
- Perplexity: Perplexity crawlers
- Cloudflare: Block AI bots
- Ahrefs: An analysis of AI Overview brand visibility factors (75K brands studied)
- Google Search Essentials: spam policies
- Google Search Central: Evaluating third-party SEO advice