About six months ago I did something I should have done years earlier. I opened Claude and asked it to find every keyword a Sydney business owner might search when they need what we offer. What came back stopped me cold.
Forty-plus search terms, real phrases people type into Google every month, where Gibson had zero presence. Competitors I know are weaker on delivery were sitting in positions one through five. We were nowhere. And I've been in business since 2006.
That afternoon I ran the same process for three clients. A property manager in Parramatta not showing up for suburb-level searches. A plumber in Penrith missing every emergency search within 10 kilometres of his depot. A medical practice invisible for the exact procedure names patients search before booking.
The keywords existed. The volume was there. Nobody had found them, because nobody had looked this way. That's when AI changed everything.
Why traditional keyword research fails most Sydney businesses
Generic keyword research is designed for brands with domain authority, budget for link building, and patience measured in years. A Sydney electrician or a Parramatta real estate principal doesn't have any of those in surplus.
Lower volume, higher intent.A keyword with 90 searches a month in a specific Sydney suburb converts far better than a national term with 9,000. The person searching “emergency electrician Blacktown Saturday” has already decided they are hiring. Local specificity as the moat. The big agencies optimise for national terms, leaving a clear lane at suburb level that most small businesses never touch. AI as the unfair advantage. What used to require a specialist and a full tool suite now requires a well-structured prompt and 20 minutes.
Gibson's 6-step AI keyword research process
01
Pick Your Seed
One keyword, one problem
02
AI Expansion
Claude finds 80+ variations
03
Intent Filter
Info vs Transactional
04
SERP Check
Can you actually rank?
05
Local Layer
Suburb-level keywords
06
Map & Track
GSC + monthly review
Step 1: Pick your seed keyword
A seed keyword is where you start. Not your business name. Not your industry. The actual phrase your best client types when they have the problem you solve.
- For a Sydney plumber: blocked drain
- For a Parramatta property manager: property management fees
- For a trades marketing company: lead generation for tradies Sydney
The seed keyword test: would your best client type this into Google at the moment they are ready to spend money? If yes, it is a seed.
Step 2: AI expansion — the part that changes everything
Open Claude, ChatGPT, or Perplexity. Use this prompt structure, adapted to your business:
You are a homeowner in Western Sydney who needs an electrician. You are comparing options before booking.
List every phrase you would type into Google across three stages: when you first notice the problem, when you are comparing options, and when you are ready to call.
Include suburb-level variations for Blacktown, Penrith, Parramatta, Merrylands, Wentworthville. Write these the way a real person speaks, not industry terminology.
What comes back is not generic. It is the actual language your customers use, organised by where they are in the buying decision, with local geography already built in. For the electrician example, Claude returned phrases across three categories:
Step 3: Intent filter
Every keyword belongs in one of four buckets: informational (they want to learn), navigational (they want a specific brand), commercial (they are comparing options), or transactional (they are ready to book or call now). Before you do anything with a term, open Google in a private browser and look at what fills the first page.
If the first page is all blog posts, Google has decided this is informational — a service page will not crack it. If the first page is service providers with booking CTAs, Google has decided this is transactional. Match the content type to the intent every time.
Step 4: SERP check and the difficulty reality test
Run your shortlisted keywords through Google Search Console (free) or Ubersuggest's free tier. A term with 40 monthly searches in a specific Sydney suburb with clear transactional intent can be worth more than a 4,000-volume national term you will never realistically rank for. Focus on terms where the ranking pages are local businesses with limited backlink authority. That is the lane available to you.
Step 5: Local layer — where Sydney businesses win
“Property management fees” is national and competitive. “Property management fees Parramatta” is local and far more achievable. “Property management company Merrylands reviews” is hyper-local and almost entirely uncontested. The further down the geographic specificity ladder you go, the less competition and the higher the conversion intent.
Gibson Promotions — Sydney service coverage
155 Sydney suburbs. One process. Zero guesswork.
For suburb-level keyword expansion, run the same prompt with your own service and suburb list, asking Claude for transactional-only variations. The output maps directly to suburb pages or Google Business Profile service areas.
Step 6: Map and track
Every keyword that survives gets assigned to one specific URL. One keyword, one page. Two pages competing for the same term split your authority and confuse Google. Track weekly using Google Search Console — it is free, it pulls real data from Google's own index, and it shows exactly which keywords are driving impressions and clicks. Keyword research is not a one-time job. It is a monthly review and a quarterly expansion as your authority grows.
The before and after
Before
- ✕Phone stays quiet while competitors rank above you
- ✕Invisible for suburb-level searches
- ✕Ad spend on traffic you should own organically
- ✕Weaker competitors outranking you on Google
After
- ✓Phone rings from suburb-level searches
- ✓Ranking for terms you weren't targeting
- ✓Showing up in AI-generated answers
- ✓Organic traffic compounding monthly
Prompt research: the layer most businesses miss
Traditional keyword research captures what people type into Google. Prompt research captures what people ask AI tools like ChatGPT, Perplexity, and Claude.
A Google search: property manager Parramatta
The same intent in ChatGPT: “I own an investment property in Parramatta and I'm not happy with my current manager. What should I look for when switching, and what questions should I ask before signing a new agreement?”
If your website answers the Google query, you might rank. If it answers the ChatGPT query, you get cited in AI responses and reach prospects who never go to Google at all. Take your top five commercial keywords and expand each into three full-sentence questions a real customer would ask an AI assistant. Then check whether your existing content actually answers those questions completely.
You can check how your site is performing in AI-generated answers right now with the free AEO Readiness Checker. It takes 60 seconds and shows you exactly what is holding your site back from appearing in ChatGPT, Perplexity, and Google AI Overviews.
Worked example: Sydney electrician, Western Sydney
Seed keyword: electrician Blacktown. After running the AI expansion prompt for transaction-intent terms only, Claude returned:
- emergency electrician Blacktown
- licensed electrician Blacktown same day
- electrician Seven Hills call out fee
- after hours electrician Quakers Hill
- safety switch repair Blacktown cost
- smoke alarm installation Blacktown price
SERP check confirms: page one for “emergency electrician Blacktown” shows local businesses with limited backlink authority. No national chains dominating. Achievable for a local electrician with a properly optimised Google Business Profile and a dedicated service page.
Mapped to /electrician-blacktown/ as the primary page, with suburb extensions built out from there for Seven Hills, Quakers Hill, and Rooty Hill. Total time to run this process from scratch: 25 minutes using Claude. Cost beyond a Claude subscription: zero.