LLM SEO for ChatGPT, Gemini and Perplexity recommendations
LLM SEO is the work of making your business easy for language models to find, read and describe correctly. ChatGPT, Gemini and Perplexity can answer a local question with a short list of names, drawn from what the model learned in training and from pages it fetches while it answers. Your website, your Business Profile, your listings and what others write about you feed both routes, and those are the parts a business can actually work on.
Why choose us for LLM SEO
We approach LLM SEO by treating your business as an entity rather than a spot in a ranking table, then working on the authority and consistency of everything published about you. That changes what the work looks like.
We structure your brand information so models can interpret who you are, what you offer and where you work, then back it with mentions on sources they trust.
Some of it sits in your hands and some doesn't. You control your website, your Google Business Profile, your directory listings and how you reply to reviews. You don't control which sources an AI product retrieves for a given question or how it words the answer, and Google's guide to generative AI search says plainly that no third-party tool has access to its internal ranking or AI systems. So the work stays on the parts that can be checked, changed and checked again.
You see the picture as the work progresses. We track where your business appears across AI platforms and report it plainly, so you always know what your spend is doing.
What's included
Entity optimisation
Your business details, services and service areas aligned across the web so language models read one clear, consistent entity. Google's Business Profile guidelines ask you to represent the business as it's consistently represented and recognised in the real world, with no service keywords or taglines added to the name. The work lines up your trading name, phone, address or service area, categories and service list across your site, your profile and the directories that already mention you, so no source contradicts another.
Crawler access checks
A business can be skipped simply because the bots can't reach it. The work checks robots.txt, firewall and CDN rules against the search crawlers each AI product documents: OAI-SearchBot for ChatGPT search, PerplexityBot, Claude-SearchBot and Googlebot. It also confirms your services, areas and contact details sit in the page's text rather than only inside images or scripts. Google's guidance for AI features recommends keeping important content in textual form.
High-intent topic alignment
Content connected to the questions buyers actually ask AI tools, so you surface while decisions are being made. Google describes a query fan-out technique: AI Overviews and AI Mode issue multiple related searches across subtopics to build one answer. A question about a blocked drain can spread into searches about cost drivers, call-out areas and what to check first, so the work maps those supporting questions and makes sure a page on your site answers each one.
Trusted mentions and citations
References on credible directories, publications and industry sites that models lean on when picking who to recommend. The work finds the sources an AI product already names for your kind of question, then checks whether your business is listed there accurately. Google's guide warns that seeking inauthentic mentions across the web isn't as helpful as it might seem, so the focus is real listings, memberships and local coverage, with no paid placements dressed up as editorial.
Reputation signals
Review strategy and sentiment monitoring, because how people talk about you is one of the signals we work on. In practice that means asking every customer the same way once the job is done, replying to each review, and reading which jobs and suburbs reviewers mention most. A review that names the suburb and the work done tells a reader far more than a star rating on its own.
Structured, AI-readable content
Pages organised so models can extract your services, locations and proof without guessing. That means one clear heading per service, a plain answer near the top of each section, and LocalBusiness and Service structured data where it fits. Google says structured data isn't required for its generative AI search and asks that any markup match the visible text, so the markup only restates what the page already says.
AI visibility tracking
Ongoing monitoring of where you appear in ChatGPT, Gemini and Perplexity, with the strategy adjusted as models update. Tracking uses a fixed set of buyer questions run on a schedule, each answer logged as mentioned, cited as a source, recommended or absent. Alongside it sits the AI Assistant channel in Google Analytics 4, which groups visits arriving from ChatGPT, Gemini, Copilot and similar tools.
How a language model finds and reads a business
A language model can know about your business by two routes, and LLM SEO has to cover both. The first is training: crawlers collect public pages and the model learns from that snapshot up to a cut-off date. The second is live retrieval, where the product searches the web while it answers. Google calls this retrieval-augmented generation, or grounding, and describes it as relying on its core Search ranking systems to retrieve relevant, up-to-date pages from the Search index.
The two routes explain a common complaint. An assistant answering from training alone can repeat an old phone number or a service you stopped offering, because that's what the pages said when they were collected. When it retrieves live pages, it reads what your site and listings say today. Each vendor publishes the crawlers it uses for each job, and they're worth checking against your robots.txt and firewall settings.
User-triggered fetchers sit outside the table below. ChatGPT-User and Perplexity-User visit a page when someone asks a question, and both vendors say robots.txt rules may not apply because a person requested the visit. Blocking training crawlers is a business decision the table can inform. Blocking the search crawlers works against the whole point of LLM SEO.
| Product | Crawler for search answers | Crawler for model training | What the vendor says about blocking |
|---|---|---|---|
| ChatGPT | OAI-SearchBot | GPTBot | Sites that opt out of OAI-SearchBot won't be shown in ChatGPT search answers, though they can still appear as navigational links |
| Google AI Overviews and AI Mode | Googlebot, through the Google Search index | Google-Extended token for Gemini training and grounding | A page must be indexed and eligible for a snippet to appear; Google-Extended doesn't affect Search inclusion or ranking |
| Perplexity | PerplexityBot | Not documented; Perplexity says its crawlers aren't used to collect content for training foundation models | Perplexity recommends allowing PerplexityBot so a site can appear in its search results |
| Claude | Claude-SearchBot | ClaudeBot | Blocking Claude-SearchBot may reduce a site's visibility and accuracy in user search results |
LLM SEO and traditional SEO side by side
The two share most of their foundations and differ in what counts as a win. Google states there are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimisations necessary: a page must be indexed and eligible to show in Search with a snippet. That makes ordinary SEO the entry ticket for Google's AI features. Assistants outside Google run their own retrieval, which is where listings, mentions and crawler access start to matter separately.
A service page that states its facts plainly can serve both surfaces, and a Business Profile with the wrong hours hurts both. Our AI search optimisation page covers the Google side in more depth.
| Question | Traditional SEO | LLM SEO |
|---|---|---|
| What counts as a win | A higher position for a page in a results list | Your business named, cited or recommended inside a generated answer |
| What the searcher sees | A list of links, a map pack and ads | A written answer with a few names and a handful of source links |
| Unit of work | A page and the query it targets | The business as an entity, across every source that describes it |
| Where the evidence lives | Your site and the links pointing to it | Your site plus profiles, directories, reviews and coverage the answer draws on |
| How it's measured | Rankings, clicks and conversions in Search Console and Analytics | A repeated question set logged by product, plus AI referral visits and Google's generative AI impressions |
| What you can't control | Google's ranking systems | Which sources a product retrieves and how it words the answer |
How an LLM SEO engagement runs
An engagement usually starts with measurement and access, because there's no point writing new pages for a crawler that can't reach the site. The sequence below is the usual order, and how long each step takes depends on how many listings and pages need fixing.
Some things stay out of it on purpose: bought reviews, pages generated in bulk with nothing new on them, paid listings passed off as editorial, and any promise that a product will name you.
- Set the baselineWrite a fixed list of the questions your buyers ask, by service and area, and run it in ChatGPT, Gemini, Perplexity and Google AI Mode. Log each answer as mentioned, cited, recommended or absent, with the sources the product named instead.
- Check accessReview robots.txt, firewall and CDN rules, confirm key pages are indexed in Google, and confirm services and areas sit in the page text.
- Clean up the entityMatch name, contact details, categories, hours and service areas across the website, Google Business Profile and directories. Fix or redirect old pages that still describe a former address or service.
- Map content to the questionsTurn the baseline's supporting questions into page sections or new pages: what affects cost, how to choose, which areas you cover, what to do before the tradie arrives.
- Build citable proofPublish facts a model can quote: the job types you take and decline, licence details where the trade requires them, named authors and dated local FAQs. Our guide to creating evidence AI can cite goes further.
- Recheck and reportRun the same question set on a schedule, compare it with the baseline and note which sources changed.
A worked example: logging AI visibility for one trade
Say an electrician in Nambour wants to know whether AI tools mention the business for switchboard upgrades and EV charger installs. This example is hypothetical, and every entry in the table is illustrative, not a benchmark.
The first check runs a short set of questions worded the way a homeowner would type them, such as 'who installs EV chargers near Nambour' and 'do I need a switchboard upgrade for an EV charger'. The log records what each product did with the business and which sources it named instead.
Read across the rows and the jobs are clear. The old phone number traces back to a directory listing from before the business changed numbers, so it gets corrected at the source. The EV charger page answers the technical question but never names the suburbs it covers, so a service-area section goes on. The same questions run again the next month, and the report shows what changed.
Referral traffic gets its own check. Google Analytics 4 groups visits from ChatGPT, Gemini, Copilot and similar assistants under an AI Assistant channel, which excludes Google's AI Overviews and AI Mode. Search Console's generative AI performance report shows impressions of your links in AI Overviews and AI Mode, impressions rather than clicks. Neither tool shows an answer that named you without a link, which is why the question log stays. More in how to measure ChatGPT referrals.
| Question | ChatGPT | Perplexity | Google AI Mode | Sources named instead |
|---|---|---|---|---|
| EV charger installer near Nambour | Absent | Mentioned | Absent | Two directories and a competitor's service page |
| Switchboard upgrade for an EV charger | Absent | Cited (service page) | Cited (service page) | An energy retailer's guide |
| Electrician open Saturday in Nambour | Mentioned, old phone number | Absent | Absent | An outdated directory listing |
When an AI answer gets your business wrong
Wrong answers happen, and Google's own help page says AI Overviews can and will make mistakes. Typical errors for a local business include an old address, a service you've dropped, a suburb you don't cover, or details blended with a similarly named competitor. The fix starts at the source, since the answer is repeating what it found or learned.
Open the links the answer cites and find the page carrying the wrong detail. Correct it if it's yours, ask the directory or publisher to update it if it isn't, and redirect retired pages on your own site to their current equivalent. Then use the product's feedback tools; AI Overviews has a thumbs-down icon with a Report a problem option. Recheck the same question a few weeks later.
A product answering from trained knowledge may hold the old detail longer than one that retrieves live pages, so a correction can show up unevenly across tools. Our guide to AI source corrections walks through the follow-up, and AI answers and accurate company facts covers which facts to keep consistent.
Built for your local market
Language models answer Sunshine Coast service questions with short local shortlists. Check the source links under one of those answers and you'll often find the ordinary local sources: Business Profiles, directory listings, reviews and suburb pages that say plainly which areas a business covers.
Google's AI features guidance recommends keeping Business Profile information current. For a service-area business, current means the areas you actually cover. A business in Maroochydore that travels to Buderim, Mooloolaba and Kawana should say so the same way on its profile, its website and its main directories, because an assistant has no other way to know.
Questions we get asked
What is LLM SEO?
LLM SEO is the work of helping large language models such as ChatGPT, Gemini and Perplexity find, read and describe your business accurately. It covers crawler access, consistent business details, pages that answer buyer questions, genuine third-party mentions and regular checks of what each product says. The aim is to be named or cited inside AI answers, though no one can guarantee that a product will include you.
Is SEO dead now with AI?
No. Google says there are no additional requirements to appear in its AI Overviews or AI Mode beyond being indexed and eligible to show with a snippet, so ordinary SEO is still the way into Google's AI answers. Other assistants run search crawlers of their own, so crawlable, clearly written pages feed them too. What has changed is measurement: a mention inside an answer may never produce a click.
Can ChatGPT help with SEO?
Yes, for drafting and checking, with a person verifying the output. ChatGPT can suggest the questions buyers ask, outline a service page or compare a page against a list of facts. It can't see Google's ranking data and it can state wrong details confidently. Google's guidance says generative AI can help with research and structure, but generating many pages without adding value may breach its spam policy on scaled content abuse.
Is LLM SEO the same as GEO or AEO?
They overlap heavily. LLM SEO names the target, the language models behind AI answers. Generative engine optimisation names the answer type, and answer engine optimisation covers being the direct answer to a question, including snippets and voice results. On this site Geo SEO means something else: geographic, map-based local SEO. All of them rest on crawlable pages, accurate facts and real mentions.
Do I need an llms.txt file?
Not for Google. Its guide to generative AI search says you don't need new machine-readable files, AI text files or Markdown to appear in Search, including its AI features, because Google Search doesn't use them. The crawler pages from OpenAI, Perplexity and Anthropic describe robots.txt controls rather than llms.txt. Adding one does little harm, but it never replaces crawlable pages with the facts in plain text.
How long does LLM SEO take?
There's no fixed timeline. Access fixes can register fast: OpenAI says its systems can take about 24 hours to adjust after a robots.txt change. Corrected listings show up when a product next retrieves them, while answers drawn from trained knowledge change only when a model is updated. A sensible checkpoint is rerunning the baseline questions monthly and judging the trend over several cycles, never a single check.
Will my business stay recommended?
Not automatically. AI answers vary between runs and shift as products change and other businesses improve their own sources. Holding a place means keeping your details current everywhere, adding new proof as jobs are completed, and rerunning the same questions so a drop shows up early. If a competitor starts appearing where you were, the source column in the log usually shows which page or listing put them there.
Keep exploring
Want the phones ringing?
Tell us your suburb and your trade. We'll show you which searches and suburbs we'd target first.