Being on page one of Google is no longer the goal, because a growing number of your customers never see page one. They ask an AI for a shortlist and they ring someone on it.
This is the 6-step version of getting onto that shortlist, in order. The first step is free and takes 5 minutes, and it is most of the value. Prompts are written out throughout.
How people search now
Someone looking for what you sell used to type 3 words into Google and get 10 blue links. They scrolled, they clicked, they compared. Every piece of SEO advice ever written rests on one assumption: the person will click.
That assumption is breaking. They now describe their whole situation to ChatGPT or Claude in 4 sentences, and get back an answer naming 2 or 3 businesses. Then they stop. There is no page 2 to scroll to, and most AI conversations end without anyone visiting a website at all.
So the prize changed. It used to be a link. Now it is being one of the names inside the answer. The jargon for this is AEO, answer engine optimisation, or just AI SEO. The name does not matter. The shift does.
2 numbers worth knowing, and we have linked the originals so you can check them yourself rather than take our word for it:
- G2 surveyed 1,076 business software buyers in March 2026 and found 51 per cent now start their research in an AI chatbot rather than a search engine. A year earlier that was 29 per cent. Honest caveat: those are people buying software, not people buying candles or booking a physio. Treat it as the direction of travel, not your exact market.
- Ahrefs published their own website's numbers: AI search sent them 0.5 per cent of their traffic, and that sliver produced 12.1 per cent of their signups. Again, honestly: that is one company's data, and it is an SEO company whose audience is unusually AI-heavy. But the shape of it matches what everyone is seeing.
That second number is the good news, and it is the reason this is worth your Tuesday evening. AI traffic is tiny and converts far better than search traffic. It makes sense when you picture what the person just did: they described their exact situation, got a shortlist, compared the options, and read about the weaknesses. By the time they reach you they are not browsing. They are deciding.
Which means a small business named in 20 answers can beat a big one with 100,000 random visitors. And big companies are slow at this. 4 people have to approve a page that names a competitor, so most of them will never publish one. You can publish one this week. Speed is the only advantage you need here.
Step 1: Decide the one thing you want to be known for
Start here because it is free, it takes 5 minutes, and everything else is this one sentence repeated in more places.
Most business bios describe what the business is. "Family owned since 2009." That tells a machine almost nothing about when to recommend you. What works is a specific claim about who you help and what you do for them, with your proof stacked behind it:
[the one thing] + [proof] + [proof] + [proof]
So not "boutique fitness studio". Instead "strength training for women over 40 in Newcastle", followed by the years running, the number of members, the qualification. The first version tells an AI what you are. The second tells it what to recommend you for.
Picking that sentence is the hard part, so let AI do the first pass:
Here is my business: [describe what you do, who you serve, where you are, and how long you have been going]. Suggest 5 options for the one thing I could realistically be known for. Each should be a specific claim about the value I provide to a specific group of people, not a description of my industry. For each option, list which of my actual credentials would prove it, and tell me which claim I have the most evidence for right now.
Write it in normal words. Nobody searches for "bespoke wellness solutions".
Step 2: Say the same thing everywhere
This is the part almost everyone gets wrong, and it is the actual mechanism.
An AI is not looking your business up in a database. It is noticing which words keep appearing next to your name across the whole internet. If your name sits near "strength training for women over 40 in Newcastle" on 30 different pages, it treats that as true and repeats it. If your website says one thing, your Instagram bio says another, and a directory listing from 2021 says a third, it has nothing confident to repeat. So it names someone else.
Consistency is not tidiness here. It is the entire game.
Before you paste it anywhere, check one thing. Some website hosts block AI crawlers by default, and if yours does, everything you do on your own site is invisible. Look at yoursite.com.au/robots.txt and make sure these 4 are not disallowed:
OAI-SearchBotfor ChatGPTClaude-SearchBotfor ClaudePerplexityBotfor PerplexityGooglebotfor Google's AI answers
Then put a version of your claim into every surface you own: your homepage and about page, every social bio, your Google Business Profile (this one matters more than LinkedIn if you serve a local area), your email signature, and any directory or marketplace that lists businesses like yours.
Review sites deserve their own line. In that same G2 study, review sites were the second most used source after AI chatbots. They are free, you just have to fill in the form, and while you are there, ask 5 happy customers for a review. Real reviews with specific details get quoted by AI far more often than a star rating does.
Here is the one thing I want to be known for: [paste your claim]. Here are the descriptions of my business currently live on my website, Google Business Profile, Instagram and [other profiles]: [paste each one]. Point out every place these contradict each other or describe me differently. Then rewrite each one so they all make the same core claim, adjusted for that platform's length limit and tone. Keep the exact same key phrase in all of them.
Step 3: Write the pages AI actually quotes
3 formats get cited far more than ordinary blog posts, because they answer the question the person actually asked:
- The "best X" list. "Best strength training studios in Newcastle." A ranked list of the real options.
- The "X vs Y" comparison. Head to head, honestly.
- The "X alternatives" page. For people who already have someone and are unhappy.
Someone asking "what is the best X for Y" is asking a comparison question, and an AI answering it wants a source that already did the comparing. Give it one.
And yes, you can write these about your own category and include yourself. Everyone in the top 3 answers is doing exactly that. The rules if you do: be honest, and if a competitor is genuinely better at something, say so. Include the ones that genuinely belong, even the strong ones. Give real prices, not "affordable". A page where you win everything reads like an advertisement, and models are getting good at spotting those.
Step 4: Write so any one paragraph can stand alone
An AI does not quote your page. It grabs one passage and drops it into an answer with none of the surrounding context. So every chunk has to survive on its own.
5 changes, easiest first:
- Make headings into real questions. Not "Pricing". Instead "How much does a session cost?" That matches how people ask.
- Answer in the first 2 sentences under each heading. Do not build up to it. Say the answer, then explain.
- Trade every vague word for a specific one. "$85 a session" gets quoted. "Affordable" never does. "Cuts your admin from 3 hours to 20 minutes" beats "saves time".
- Say where your information came from. One line like "prices checked on each studio's own website in August 2026" does real work, because models weigh visible sourcing.
- Keep the same rule in any table or list. A row saying "great for teams" is wasted. "Teams of 5 to 20 across 2 locations" can be lifted straight into an answer.
The test: cover the whole page except one paragraph. Can that paragraph answer a question by itself? If not, rewrite it.
Here is a page from my website: [paste the text]. Rewrite it so an AI can quote any single paragraph out of context. Turn the headings into the questions people actually ask. Put a direct answer in the first 2 sentences under each heading. Replace every vague phrase with a specific number or fact, and flag anything I need to fill in myself. Use Australian English and no em-dashes.
2 supporting jobs while you are in there. Link your own pages to each other, 5 or more links per important page. And re-check what you have already published every 3 months: verify the prices, update the year, remove anything that has closed. A page that already has trust flips into AI answers in weeks, where a brand new page starts from zero.
Step 5: Get named on other people's sites
Here is the uncomfortable one. Around 85 per cent of the brand mentions AI makes come from sites that are not yours, based on an analysis of 21,311 mentions. Everything on your own website has a ceiling until other people start talking about you.
The free version takes an afternoon:
- Ask ChatGPT, Claude and Perplexity your main category question: "what are the best [what you do] in [your area]?"
- Write down every article they cite. Those are the pages the AI trusts in your category.
- Open each one. Are you on it?
- If not, email the author. Short, specific, no pitch: you read their piece, you run [your thing], here is the one thing you do that the others on their list do not, happy to send photos, honest notes and where you are weak if it helps the next update.
Some will ignore you. Some will add you. Every one that adds you becomes a source an AI can quote.
Beyond that sits the advanced tier, in rough order of effort: a wire service press release when you launch something genuinely notable, and guesting on podcasts or YouTube channels in your space. A podcast episode is worth more than it looks, because the show notes page sits on someone else's trusted domain with your name on it permanently, and gets re-crawled every month. 12 guest spots build 12 permanent pages.
Step 6: Track whether it is working
Your scoreboard is no longer traffic. It is how often you get named, and you count it by hand.
Once a month, take 10 questions your customers would realistically ask, paste each into ChatGPT, Claude, Perplexity and Google, and log whether you were named and who else was. That spreadsheet is your ranking report now. Then watch 2 supporting numbers: how many people search for your business name (Google Search Console shows this), and how many enquiries or sales you actually get.
I run [what you do] for [who] in [where]. Write the 10 questions my customers would most likely type into ChatGPT when they are ready to choose someone like me. Make them realistic and detailed, the way a real person describes their situation, not short keywords. Then build me a tracking table with columns for the question, the date, whether my business was named, which competitors were named, and which sources the answer cited.
If you want this to run itself rather than living in your head, the rules can go into a file the AI reads every time. That is what a skill is, and we cover it in skills, commands and workflows.
How long this takes to show up
Not next week. This is compounding work, and how fast it moves depends on how competitive your category is and what you already had going in. Anyone promising you results in 7 days is selling something.
What you get is a version of your business that the internet agrees on. Because a model does not rank your website. It reads your site, your profiles, other people's lists, a review, a podcast page, and it looks for the version of you that shows up most often. Then it repeats that version.
The winner is not whoever writes the smartest page. It is whoever is described the same way in the most places. That is why a 5 minute bio fix belongs in the same guide as a press release. They are the same move at different sizes.
Common mistakes
- Waiting until you understand it fully. The 5 minute version in step 1 is most of the value. The people winning at this are not smarter than you, they just started while it still felt too early to matter.
- Describing yourself 4 different ways. The single most common failure, and the cheapest to fix. Fuzzy means an AI will not risk recommending you.
- Writing "what is a [your category]" explainer posts. Those bring in students and researchers, and an AI answers them completely without naming anyone. Write for people who are ready to choose.
- Only working on your own website. Most of what an AI says about you comes from somewhere else. Your own site has a ceiling.
- Measuring traffic. If you judge this by visitor numbers you will conclude it failed. Count mentions and enquiries.