Category: Uncategorized

If you’ve asked ChatGPT, Perplexity, or Gemini who they’d recommend in your category and your name didn’t come up, you’re not alone. Most Australian businesses — even successful, well-established ones — are invisible to the AI systems their customers are increasingly using to find them.

The numbers are stark: one in three consumers now starts product research in ChatGPT, Perplexity, or Google AI Overviews rather than a traditional search bar. AI referral traffic grew 25 times in a single year. And 58.5% of Google searches already produce zero clicks to any website. If you are not being cited by the AI, you are missing buyers who never make it to Google at all.

Here’s why, and what actually moves the needle.

The models don’t know you exist — or they’re confused about who you are

The most common cause of AI invisibility isn’t that your business is bad. It’s that the models don’t have confident information about who you are, what you do, and where you do it. This is an entity problem, not a content problem.

Think of it like this: the models maintain a rough mental model of every entity they’ve encountered — businesses, people, places. If your entity is fuzzy (inconsistent name across directories, no structured data, no external corroboration), the model’s confidence in recommending you is low. Low confidence = no citation.

The three signals that actually drive recommendations

Entity confidence. The model needs to be certain of your name, location, category, and what makes you the right option for a specific type of buyer. This is solved by schema markup, NAP consistency across the web, and a knowledge-graph entity.

Third-party corroboration. The model needs to see your name in sources it trusts: trade press, industry directories, podcast transcripts, local government sites, community forums. Not links — mentions. Unlinked brand mentions in authoritative sources are a primary signal for AI recommendation systems in 2026.

Topical authority. For the specific questions buyers ask the AI in your category, your website needs to be the most comprehensive, structured answer. Not the most content. The most useful answers to the exact queries — written in a format the retrieval pipeline can act on.

What doesn’t work

More blog posts. More keywords. More backlinks. These are the levers of traditional SEO, and while they’re not harmful, they don’t move the AI visibility needle in any meaningful way. We’ve audited businesses with 800 indexed posts and a Visibility Index of 11. Content volume is not the same as retrieval authority.

What the timeline looks like

Run the free scan above. If you’re under 50, the most common path is:

• Days 1–14: Schema, entity consolidation, NAP reconciliation. Typically moves the index 15–25 points. The infrastructure work.
• Days 15–60: Authority content — structured answers to the specific buyer queries in your category. FAQPage schema, definitional pages, comparison content. The depth work.
• Days 30–90: Citation engineering — earned mentions in trade press, industry directories, podcasts, community sources. The corroboration work.

First citations in Perplexity and ChatGPT typically land inside two weeks of the schema work going live. The index usually crosses 50 by day 60, and 65 by day 90. The compounding moat forms between months 4 and 6.

The window is closing

Every category has a window where establishing AI visibility is relatively cheap. In some — accounting, legal, trade services — that window is still wide. In others — digital marketing, real estate, financial planning — the early movers are already dominant and closing the gap is expensive. The best time to run the scan was six months ago. The second best time is today.

If you’ve run the Befound scan and you’re looking at a number between 0 and 100, here’s what you’re actually looking at — and what to do with it.

How the score is calculated

The Visibility Index is an aggregate score across four AI systems — ChatGPT, Perplexity, Gemini and Claude — for the twelve highest-intent buyer queries in your category. We’re not measuring whether your website loads fast or whether your meta descriptions are compelling. We’re measuring whether the AI your customers are already using would recommend you by name.

For each query, we look at three things: does the model mention your business by name? Does it recommend you specifically? And does it cite a source connected to you? Each answer is weighted and aggregated into the score you see.

What the bands mean

0–25: Invisible. The models either don’t know your business exists, or they know the category and consistently name competitors instead. This is the most common starting position for businesses that have been online for years but haven’t structured their presence for retrieval.

26–45: Emerging. The models have partial entity recognition — they might know your name in one or two contexts, or one model picks you up while others don’t. This usually means your on-site entity signals are partial and your third-party corroboration is thin.

46–65: Mixed. You’re in the conversation, but not reliably. The models cite you for some query types and miss you for others. In practice, this means you’re winning some AI-referred inquiries but losing most of them — usually to a competitor who got their entity and citation infrastructure right first.

66–80: Competitive. You’re a named option in most relevant AI answers. You’re not necessarily the default recommendation, but a buyer who asks broadly will encounter your name. At this level, the work shifts from getting recognised to getting preferred.

81–100: Dominant. You’re the default recommendation in your category or sub-category. This is a durable moat — it takes competitors 6–12 months of sustained work to close a 20-point gap — but it’s not permanent without maintenance.

The single number that matters most inside the report

The per-model breakdown matters more than the aggregate. A score of 48 built from ChatGPT:62, Perplexity:58, Gemini:31, Claude:29 tells a very different story than 48 built from four scores clustered around 48. In the first case, you’re winning web-retrieval-heavy models and losing training-data-heavy ones — the fix is earned third-party mentions, not more on-site content. In the second, you have a broad but shallow presence — the fix is depth of topical authority.

What happens next

If you’re under 50, the Report is the right move. It maps every query in your category across every model, identifies the specific gaps — entity confidence, NAP consistency, schema, third-party corroboration — and gives you a 90-day roadmap your team can execute or hand to us.

If you’re over 50 and want to move fast, the Programme is the shortest path. We ship the citation engineering, schema, and authority content. Most Programme clients cross 65 inside 90 days.

Either way: the scan is the start, not the finish. The number tells you where you are. The report tells you why. The programme gets you where you need to be.

NAP stands for Name, Address, Phone. It is the most unglamorous concept in local digital marketing; and it is responsible for more AI search invisibility than any other single factor. In our audit portfolio, 94% of local businesses have at least three NAP inconsistencies across their web presence. Most have more than ten.

Why NAP matters more for AI than it did for Google

Google local ranking algorithm uses NAP consistency as one signal among many. Its tolerance for minor inconsistencies is reasonably high; Google has been dealing with messy local business data for 20 years and has built heuristics to handle it. AI retrieval systems have not developed the same tolerance. They treat inconsistent entity data as evidence that they are dealing with multiple partially-overlapping entities; and they respond by reducing recommendation confidence for all of them.

The most common NAP problems we find

Legacy trading names. The business rebranded, updated the GBP and website, but left 12 directory listings with the old name. The models now have two entities with overlapping phone numbers and addresses but different names. Neither gets recommended confidently.

Address format variation. “St” versus “Street”. “Level 2” versus “L2”. “Suite 4” versus “Shop 4”. These are visually minor but semantically significant to entity-matching algorithms. Standardise to one format (we recommend using the Australia Post format for your address) and apply it everywhere.

Phone number formats. (03) 9XXX XXXX versus 03 9XXX XXXX versus +613 9XXXXXXX. Pick one national format (+61 3 9XXX XXXX is generally best) and use it consistently across all digital properties.

Orphaned directory listings. Businesses that have been running for more than five years almost always have listings on directories they never set up; scraped from other sources and never claimed. These often have outdated information. We find an average of eight unclaimed listings per business audit.

How to audit it yourself

Search for your exact business name in quotes on Google. Click through the first 20 results. Note every variation of your name, address, and phone number you find. That is your NAP audit. Then do the same on Bing, because some AI systems weight Bing Places data.

Prioritise fixing: Google Business Profile, your own website, Yelp, True Local, Yellow Pages, Hotfrog, Cylex, and any industry-specific directories relevant to your category. That covers roughly 80% of the citation weight.

How long does the fix take?

For most local businesses, a thorough NAP reconciliation takes 4 to 8 hours of actual work spread across 2 to 3 weeks. It is not glamorous. It does not produce impressive-looking output. But in our portfolio, NAP reconciliation alone moves the Visibility Index an average of 18 points; more than any equivalently-sized content investment. It is the foundation that makes everything else work properly. Do it before you publish new content. Do it before you run a citation campaign.

Anonymised. Drawn from two regional allied-health programmes.

Background

A physiotherapy practice in Ballarat (three practitioners, established 2016, bulk-billing and private-fee services) was invisible in AI-generated answers for “physiotherapist in Ballarat” and its variants across all four major AI systems. The practice had a functional website, 140+ Google reviews averaging 4.8 stars, and a reasonable local SEO ranking. Visibility Index at audit: 16 out of 100.

What the audit revealed

Three structural problems, each fixable. No schema on the website: the WordPress site had no MedicalBusiness or Physician schema; the models had no structured signal confirming the business type, the practitioners credentials, or the conditions treated. One model was categorising the practice as a “wellness centre” rather than a physiotherapy clinic, which affected which queries it appeared in.

Practitioner profiles were non-existent. In allied health, the practitioners are the entity as much as the practice. A physiotherapy clinic with no structured profiles for its physios (no AHPRA registration numbers in schema, no authored content, no Person schema) is invisible in the “best physio for condition” queries that account for roughly 60% of the category high-intent search volume.

Third-party mentions were thin and generic: two health directories, both with generic entries that mentioned “physiotherapy” but no specific conditions, techniques, or service differentiators.

The programme; 11 weeks

Weeks 1 to 2: Schema foundation. Added MedicalBusiness schema with medicalSpecialty, availableService, and priceRange. Added Person schema for all three practitioners with AHPRA registration numbers, credential lists, and sameAs links to their AHPRA public register entries. This alone moved the index from 16 to 34.

Weeks 2 to 5: Condition-specific content. Published six condition pages (lower back pain, sports injuries, post-surgical rehab, neck pain, pelvic floor, dry needling) each written with a specific structure: condition definition, how physiotherapy addresses it, what to expect in sessions, approximate duration of treatment, and explicit FAQ schema addressing the questions buyers ask AI assistants. Average length: 1,600 words. Each page authored and schema-attributed to a specific practitioner.

Weeks 4 to 8: Citation engineering. Placed practitioner profiles in the Australian Physiotherapy Association member directory (previously unlisted). Secured two editorial placements in the Ballarat Courier health section. Earned a mention in a local community Facebook group thread about post-natal physio. Added to the Better Health Channel regional provider search.

Weeks 6 to 11: Review structure. Added AggregateRating schema to the homepage pulling from Google review data. Launched a structured post-appointment review request SMS that directed patients to leave condition-specific reviews mentioning the treatment type; not just stars, but context. “Great help with my back pain after the marathon” is a more useful retrieval signal than “5 stars, highly recommend”.

Results at week 11

• Visibility Index: 16 to 71
• ChatGPT: named in 9 of 12 priority queries (from 0)
• Perplexity: named in 11 of 12 priority queries (from 2)
• New patient enquiries via website: up 41% week-on-week in weeks 9 to 11 versus prior period
• Specific condition queries (“physio for lower back pain Ballarat”): top recommendation in all four models

The learning

Allied health is a category where practitioner-level entity work unlocks category-level visibility. Adding AHPRA numbers to Person schema sounds like a minor detail; it produced an 18-point index jump in two weeks because it is a verifiable, authoritative credential signal that the models treat with high trust. In any licensed profession, the equivalent move (licence number plus credential body sameAs) is the highest-ROI single action available.

If you have been running a local business for more than five years, you have already survived one major platform shift: the move from Yellow Pages and word-of-mouth to Google. That shift took about a decade to fully play out; and the businesses that understood it early built advantages that lasted years.

The shift to AI-mediated discovery is faster; and the mechanics are different enough that the Google playbook does not transfer cleanly. Here is what has actually changed, what is the same, and what you should focus on.

What has genuinely changed

The answer surface is narrower. Google returns ten blue links. AI systems return one synthesised paragraph with two or three named options. You are either in the answer or you are not. There is no position three that still gets clicks.

The ranking signals are different. Google weights PageRank (incoming links), on-page keyword optimisation, and page experience signals. AI systems weight entity confidence (do they know unambiguously who you are?), third-party corroboration (do trusted sources mention you?), and topical depth (are you the most comprehensive, structured answer to the question?). These overlap with traditional SEO but are not the same problem.

The query format has changed. Buyers used to type two-word queries: “electrician Northcote”. They increasingly ask full questions: “who is the most reliable licensed electrician for a full house rewire in Northcote; I need someone who can work around a tenant?” The specificity of the query means the specificity of your entity signals matters more than it used to.

Citations matter more than links. In the Google era, a backlink from a quality site was the primary authority signal. In the AI era, an unlinked brand mention in a trusted source (a trade magazine article, a local council recommendation page, a Reddit comment thread, an industry association directory) is often as valuable as a backlink. The model does not crawl link graphs; it reads corroboration.

What has not changed

Being genuinely good at what you do is still the foundation. AI recommendations, like Google rankings, eventually reflect reality. Businesses with consistent reviews, genuine service depth, and real customer outcomes build sustainable AI visibility. The ones gaming signals without underlying quality will see their index decay as the models encounter contradictory signals.

Consistency matters. In the Google era, NAP consistency across directories was important for local rankings. It is equally important for AI entity confidence. The principle is the same; the reason is slightly different.

Being findable is not the same as being chosen. AI SEO gets you in the answer. Converting that appearance into an enquiry or a booking still depends on your website, your reviews, your responsiveness, and the quality of your actual offer.

The practical action right now

1. Run the free scan and get your baseline number
2. Audit your NAP consistency across all directories (we find at least three discrepancies in every business we audit)
3. Add LocalBusiness schema to your website with all relevant fields
4. Get your business into two or three industry-specific or local-authority directories you are not currently in
5. Rewrite your homepage opening paragraph so the first 80 words establish who you are, where you are, what you do, and who you serve; in plain, specific language

None of that requires an agency. It requires an hour and a willingness to treat your business web presence as a structured data problem, not just a design problem.

You have 200 five-star Google reviews. Your Google Business Profile is immaculate. You rank on the first page for your main suburb keywords. And yet when a prospective customer asks ChatGPT, Perplexity, or Claude for a recommendation in your category, your name does not appear.

This is one of the most common frustrations we hear from local business owners who have done everything right in the Google-era playbook. Here is why it happens; and what actually moves the needle for AI search.

What Google reviews signal, and to whom

Google reviews are a trust signal for Google systems: the map pack, local organic results, and to some extent Google AI Overview. They are aggregated into your Google Business Profile and contribute to Google local ranking algorithm. They do not, however, directly inform ChatGPT, Perplexity, or Claude. Those systems have different training corpora, different retrieval architectures, and different trust hierarchies.

A business with 200 Google reviews is not automatically more visible in non-Google AI answers than a business with 20 reviews; unless those reviews also exist in structured form on your own website (via AggregateRating schema) or are referenced in third-party sources those AI systems actually retrieve from.

The signals that AI systems actually weight

Entity consistency. The models need to be certain your business is a single, identifiable entity with a stable name, address, phone number, and category. Inconsistencies across directories fragment entity confidence and suppress recommendations.

Third-party corroboration. This is the big one; and it is the most different from the Google playbook. AI systems weight mentions of your business in sources they treat as authoritative: trade press, industry directories and associations, local government sites, community forums (Reddit, local Facebook groups, Nextdoor), podcasts, and niche publications. Unlinked mentions count. The model does not need a hyperlink; it needs to have seen your name in a trusted context.

Structured data on your own site. Schema markup (specifically LocalBusiness schema with correct address, areaServed, telephone, and priceRange fields) gives retrieval systems a machine-readable brief on your business. Without it, the model has to infer your entity from prose; and it frequently infers wrong.

Topical authority for your category. Content that comprehensively answers the questions buyers ask in your category (structured, specific, and honest) establishes your domain as a retrieval-worthy source. The AI does not just recommend businesses; it cites sources. If your website is a source it can cite, your recommendation rate goes up.

What to do with your reviews

Your reviews are not wasted; they need to be structured to work in AI contexts. Add AggregateRating schema to your homepage and service pages, drawing from your Google review count and average. This makes the model retrieval of your review data reliable rather than incidental. Then focus on getting review velocity across other platforms: industry-specific directories, Yelp, Hotfrog, and wherever your category buyers already look. Multi-platform review presence is a much stronger entity signal than concentration on a single platform.

Run the free AI visibility scan on this page. It takes 60 seconds and shows you exactly how ChatGPT, Perplexity, Gemini, and Claude currently see your business. The score it returns is a function of entity signals, not Google signals. It is almost always surprising; even for businesses with excellent Google presence. Especially for them.

Anonymised. Composite drawn from three regional trade-services programmes.

The situation

A family-run electrical contracting business in Geelong (two vans, four staff, established 2011) had solid Google reviews, a functioning website, and a steady flow of word-of-mouth work. When the owner started noticing that new enquiries were slowing, he ran a quick test: he asked ChatGPT and Perplexity who to call for a licensed electrician in his service area. Neither named his business. One named a Melbourne-based company with a Geelong landing page. The other named a business that had closed two years earlier.

Starting Visibility Index: 11 out of 100.

What we found in the audit

The business had three separate ABN-registered trading names appearing across different directories; a legacy of a rebrand in 2019 that was never fully cleaned up. Yelp had the old name. True Local had a different phone number. The Google Business Profile had the current name but a PO Box address while the website had the street address. From a retrieval standpoint, the models were dealing with at least three partially-overlapping entities that they could not confidently merge into a single recommendation.

Schema: none beyond the basic WordPress defaults. No LocalBusiness markup, no areaServed, no licenceNumber field (which, for licensed trades in Victoria, is a trust signal the models weight meaningfully). Third-party corroboration outside Google reviews: two citations, both from 2018.

The work; eight weeks

Weeks 1 to 2: Entity consolidation. Reconciled all directory listings to a single consistent NAP. Claimed and standardised 17 directory profiles. Added VBA licence number to the GBP, website footer, and schema. Consolidated the two old trading-name profiles with a merge request. This single step moved the Visibility Index from 11 to 29.

Weeks 2 to 4: Schema and content structure. Added LocalBusiness and Electrician schema with areaServed polygon covering all six postcodes serviced, explicit priceRange, openingHours, and paymentAccepted. Rewrote homepage opening paragraph to lead with five disambiguation signals in the first 60 words: suburb, founding year, licence number, scope of work, service area.

Weeks 3 to 6: Authority content. Published four service pages (switchboard upgrades, EV charger installation, safety switch testing, solar integration) each with 1,400 to 1,800 words, FAQPage schema, and a realistic price-range table. Price tables are consistently the most-retrieved content type for trade services; they answer the question buyers actually ask first.

Weeks 5 to 8: Citation engineering. Secured placements in the Geelong Advertiser small-business section, the local HIA chapter newsletter, and two local suburb Facebook groups (managed mentions, not ads). Added the business to the Master Electricians Australia member directory. Each placement was unlinked; the mention itself is the signal.

Results at week 10

• Visibility Index: 11 to 68
• ChatGPT: named in 8 of 12 priority queries (from 0)
• Perplexity: named in 10 of 12 priority queries (from 1)
• Phone enquiries from new customers: up 34% in weeks 8 to 10 versus same period prior year
• Of 34 new enquiries in weeks 9 to 10, 11 cited asking an AI when asked how they found the business

The learning

For established local trade businesses, the most impactful work is almost never new content. It is entity cleanup: the boring, invisible reconciliation of NAP data, schema, and directory consistency that makes the model say “I know exactly who this is and where they work.” The content work accelerates compounding. But without the entity foundation, additional content just adds more signals to an already-confused entity picture.

Most local business owners we talk to have the same reaction when AI SEO comes up: “Is this just another thing I have to pay for?” Fair question. Here is the honest answer; and the ten reasons it matters more for small and local businesses than for anyone else.

First, a quick orientation

AI SEO (sometimes called Generative Engine Optimisation, or GEO) is the practice of making your business the one that AI systems (ChatGPT, Perplexity, Gemini, Claude) recommend when someone asks a question your business should answer. It is different from traditional SEO, which aimed at Google blue links. The goal here is inclusion in the synthesised paragraph the AI writes. That paragraph might name two or three businesses. Yours needs to be one of them.

For local businesses specifically, this matters a lot. When someone new to your suburb asks who is the best dentist in Northcote, they are increasingly asking an AI, not typing into Google. The AI answer is shaped by structured data, third-party mentions, and entity confidence; not by who spent more on ads.

The ten benefits

1. You compete on merit, not budget

Google ads favour the biggest wallet. AI recommendations favour the most machine-legible entity. A three-person accounting firm with solid schema, consistent NAP, and a few earned trade-press mentions can outrank a national chain if their entity signals are cleaner. The playing field is genuinely more level.

2. The traffic that arrives is already pre-qualified

When an AI recommends your business by name in response to a specific buyer question, the person who arrives has already received a personalised endorsement. Conversion rates from AI-referred enquiries in our portfolio are 2 to 4 times higher than from organic search clicks. The AI does the persuasion; you close.

3. It compounds; and the moat is hard to close

Once the models have high entity confidence in your business, they keep recommending it. Citations beget more citations. Each mention in trade press or community forums adds corroboration weight. A local business that builds a 70/100 Visibility Index typically maintains it with far less ongoing effort than they spent getting there. Early movers in most local categories are building leads right now that will take competitors 12 to 18 months to close.

4. It works on every AI, not just Google

Your Google Business Profile helps with Google search. AI SEO helps with all four major AI systems simultaneously: ChatGPT, Perplexity, Gemini, and Claude. Each has different retrieval architecture, but the underlying signals (entity confidence, structured data, third-party corroboration) are common to all of them. One programme, four channels.

5. It fixes the invisible-in-the-next-suburb problem

Most local SEO is surprisingly poor at geographic precision. We regularly audit businesses that rank well for their immediate suburb but disappear in answers about neighbouring areas they actively service. AI SEO uses structured geographic data (areaServed polygons, not suburb-list text) that models treat as authoritative. A plumber who services eight postcodes should appear in AI answers for all eight; not just the one on their domain.

6. Reviews become machine-readable signals, not just social proof

Schema-marked AggregateRating on your site, combined with consistent review velocity across Google, Yelp, and industry-specific directories, is a retrieval signal. The AI reads the structured data and weights recent, corroborated review velocity as a trust indicator for local service businesses. Your existing happy customers are an asset you are probably under-utilising.

7. You stop losing leads to businesses that do not actually serve your area

A common pattern in local AI search: large national brands with thin geographic coverage appear in “best service in suburb” answers because their domain authority is high. Proper local entity work (specific service-area schema, local case studies, community mentions) gives the models the geographic precision to get this right.

8. Your existing content starts working harder

The blog posts and FAQs on your website probably already exist. Most are written to persuade humans, not to be retrieved by AI. Small structural changes (explicit definitional opening sentences, FAQPage schema, consistent terminology, price range tables) transform existing content from invisible to frequently-cited without rewriting from scratch.

9. It integrates with your existing Google presence

Entity signals from your Google Business Profile, schema markup, and third-party mentions all feed the same retrieval ecosystem. AI SEO is not a replacement for keeping your GBP updated and collecting reviews; it is the layer that makes all of that existing work count for AI recommendations as well as map-pack rankings.

10. The audit tells you exactly where you stand today

Unlike traditional SEO, where “results in 6 to 12 months” is the standard non-answer, AI visibility is measurable immediately. The scan on our homepage gives you a real number: where you sit against your local category, per model, today. You know the baseline before you spend a cent. For a local business owner deciding whether this is worth it, that transparency matters.

The bottom line

Most local businesses in Australia are invisible in AI-generated answers right now; not because their businesses are bad, but because the models have not been given the structured signals to recommend them confidently. The businesses that fix this in the next 12 months will own those recommendations for years.

Run the free scan. The number is honest. If it is under 50, the gap is real and closeable. If it is over 65, you are already ahead of most of your local competitors; the question is whether you want to extend that lead before they notice.