Here is a sentence we did not expect to write when we started measuring how AI recommends Australian solar installers: the most important marketing document an installer has may be their entry on an accreditation register. Across our August 2026 testing, ChatGPT mentioned accreditation in 100 per cent of its solar installer recommendations, mentioned the New Energy Tech Consumer Code in 100 per cent of them, and ran literal register lookups by business name, using site restricted searches against the register domains, before deciding who to name. A third of all its background research queries targeted an accreditation body or register.
Most installers have no idea this check is happening, and a meaningful number are failing it for reasons that have nothing to do with their actual accreditation status. This article explains the current accreditation landscape, exactly how the AI verification behaves, and the alignment work that stops you failing a check you would otherwise pass. It is the most measurable part of GEO, generative engine optimisation, which is the work of getting your solar business named by AI.
The accreditation landscape in 2026, untangled
The Australian solar accreditation system reorganised itself over recent years, and the confusion that created is now baked into AI training data, which makes stating the current facts worthwhile in itself.
Solar Accreditation Australia, SAA, operating at saaustralia.com.au, accredits the individual people who design and install systems. SAA was appointed by the Clean Energy Regulator and took over this function from the Clean Energy Council. Installer accreditation is what makes a system eligible for small scale technology certificates, so it is the accreditation with direct financial consequence for every residential job.
NETCC, the New Energy Tech Consumer Code, administered at newenergytech.org.au, approves the business that sells you the system, the retailer side, under a consumer protection code covering sales conduct, contracts and complaints. It replaced the old CEC Approved Solar Retailer program, and by late 2025 there were more than two thousand approved sellers. When state rebate schemes require an approved retailer, this is generally the approval they mean.
So the clean summary an installer should be able to recite: SAA accredits the person who installs, NETCC approves the business that sells, and the Clean Energy Council, which once ran versions of both schemes, now runs neither. That last clause matters more than it seems, because the CEC’s former accreditation domain still redirects to the CEC’s main site, and an AI system following older links to verify accreditation can land on an organisation that no longer administers the scheme. The engines are navigating this confusion in real time, and we watched ChatGPT search both the NETCC organisational domain and its consumer facing domain to triangulate.
What the AI verification actually looks like
When ChatGPT builds a solar shortlist, its visible research trail shows a consistent pattern. It discovers candidate installer names, largely from comparison sites and reviews, as we covered in the SolarQuotes analysis. Then it verifies: searches in the pattern of a register domain restricted query plus the installer’s brand name, checking whether the business it is about to recommend actually appears where its claims say it should. In our data, 53 per cent of ChatGPT’s background queries used site restricted searches, overwhelmingly against registers and industry bodies.
The three engines weight this differently. ChatGPT is the strict one: 100 per cent NETCC mention rate, 94 per cent CEC mention rate, register lookups by name. Google AI Mode mentioned accreditation in 69 per cent of answers. Perplexity was level on accreditation, at 70 per cent, but far lower on NETCC, at only 20 per cent, because it leans on comparison content rather than primary verification, a difference we unpacked in the Perplexity playbook. Across all 42 answers we collected, accreditation was mentioned in 81 per cent and warranties in 69 per cent, while licensing, interestingly, appeared in only 21 per cent. The engines have collectively decided that accreditation is the trust currency of Australian solar.
The failure mode this creates is silent and brutal. An installer who is genuinely accredited but whose register entry is under a different name than their website trades under, the company name versus the brand name, the old name after a rebrand, fails the lookup. The engine does not report the discrepancy. It simply drops the name and recommends someone whose verification succeeded. Nobody ever knows the recommendation was lost, or why.
The alignment audit: an hour that protects everything else
The fix is unglamorous and takes about an hour. Search the SAA register the way a machine would, for your installers by name. Search the NETCC approved seller list for your business by the exact name your website uses. Then check the three places the names must agree: your register entries, your website footer and about page, and your Google Business Profile and comparison site profiles. Every mismatch, trading name versus entity name, missing branch locations, a director’s name on the accreditation while the brand name is on the van, is a lookup that can fail.
Then put the facts where the engines read them: your NETCC approval and your installers’ SAA accreditation stated in plain text on your site, with the names exactly as registered, ideally on the homepage and about page rather than buried on a compliance page. Add them to your schema and llms.txt. This is the solar specific version of the six fact publishing we found missing on most Australian homepages in our 500 homepage study, and for solar the accreditation facts outrank the others, because we have measured the engines checking them specifically.
While you are there, state your warranty position with the same precision, workmanship warranty years, panel and inverter product warranties for the brands you fit, because warranty mentions appeared in 69 per cent of answers and warranty specifics are exactly the kind of checkable claim that survives synthesis into an answer.
Why this beats most marketing spend
Consider the asymmetry. The verification layer costs nothing to pass, an hour of alignment and some plain text publishing, and failing it silently nullifies every dollar spent on the discovery layer: the comparison profiles, the reviews, the content, the ads. It is the narrowest bottleneck in the funnel and the cheapest to fix. It also compounds with the compliance work you already do: the same accuracy that satisfies CEC and ACCC advertising rules is what makes AI verification succeed, so the trades practices version of honesty and the AI visibility version of honesty are the same document set.
And it is defensible in a way content is not. Competitors can copy your blog in an afternoon. They cannot copy your years of consistent register presence, review accumulation and name alignment, which is why verification heavy engines like ChatGPT produce more stable recommendations than the volatile surfaces, and why the installers who pass verification keep getting named while the answers around them churn, the churn we measured in the volatility study.
The register pattern beyond solar
Solar is the clearest measured case, but the behaviour generalises, and if you operate in any regulated Australian industry the same check is running on you. Across our studies, ChatGPT’s background research consistently included site restricted searches against official sources, government and industry domains took 12 per cent of its citations against 1 per cent for Perplexity, and the pattern of its lookups, register name plus business name, repeated across categories. We watched versions of the same verification when studying builders, where practitioner registration was the filter, and lawyers, where it reached for professional directories and regulator sites, both documented in how AI recommends builders and how law firms get recommended.
The general rule for any licensed or accredited business: identify every public register your business appears on, licence boards, accreditation schemes, professional memberships, and treat your entries there as marketing surfaces with exactly one job, matching your public trading identity precisely. Then state each credential on your own site in plain text with the registered name and number, so the claim and the register confirm each other in one lookup.
There is also a defensive version of this insight worth a paragraph. Because engines verify against registers, lapsed or inconsistent entries do not just fail to help, they actively contradict your website’s claims in the exact place machines check. An accreditation your site still claims after it lapsed, a branch office on your website that no register knows about, an old entity name that never got updated after a restructure: each is a discrepancy an engine can find in one query, and the safe assumption is that eventually one will. The businesses treating register hygiene as an annual compliance chore are maintaining, without knowing it, the single data source AI trusts most about them.
Solar happens to be the industry where we can prove all this with percentages. If you are reading this from plumbing, building, law, finance or health, the sensible move is to assume the percentages apply to you and spend the same hour on alignment, because the cost of being wrong about that assumption is an hour, and the cost of the other wrong assumption is every recommendation you never knew you lost.
If you want your register entries, website and profiles checked against what ChatGPT verifies, that alignment audit is where our GEO work for solar installers starts.
Frequently asked questions
What is the difference between SAA and NETCC?
Solar Accreditation Australia accredits the individual installers and designers who do the work, which is required for STC eligibility, and took this function over from the Clean Energy Council. NETCC approves the retail business selling the system under a consumer code, and replaced the old CEC Approved Solar Retailer program. Person versus business, installation versus sale.
Does ChatGPT really check accreditation registers?
Measurably, yes. In our August 2026 testing ChatGPT mentioned NETCC in 100 per cent of solar recommendation answers, a third of its background research queries targeted an accreditation body or register, and 53 per cent of its lookups used site restricted searches, including literal register domain plus business name checks on installers it was considering.
I am accredited but AI never recommends me. Why?
The most common cause we find is name misalignment: the register holds your entity name or an old trading name while your website and profiles use your brand name, so the machine lookup fails and you are dropped silently. Audit the exact strings on the SAA and NETCC registers against your website and profiles, and align them everywhere you control.
Is CEC accreditation still a thing?
The Clean Energy Council no longer runs installer accreditation, which moved to SAA, or retailer approval, which moved to NETCC. CEC membership still exists and engines still mention the CEC, partly on older data. State your current SAA and NETCC standing explicitly so answers about you rest on the current scheme rather than the historical one.
Where should accreditation appear on my website?
In plain text, names exactly as registered, on your homepage and about page, echoed in your schema and llms.txt, not solely as a logo image or a PDF certificate. Machines quote and verify text. A logo without matching text is invisible to the check that decides whether you get recommended.
We are rebranding. What does this mean for our accreditation entries?
Treat the registers as part of the rebrand project plan, not an afterthought. Update SAA and NETCC entries, your website, your Google profile and your comparison platform profiles in the same fortnight, because every week they disagree is a week the verification lookup can fail against whichever name the assistant tries first. Keep a line on your about page noting the former name for a year or so, formerly trading as, since older reviews, threads and citations under the old name are evidence you want connected to the new one rather than orphaned.
Does any of this matter for Perplexity and Google AI Mode?
Less, and that is worth knowing. Perplexity mentioned NETCC in only 20 per cent of answers and leans on comparison content instead; AI Mode sits in between. The alignment work is aimed primarily at ChatGPT, which is both the strictest verifier and, for many buyers, the assistant they actually ask. The other engines are won mainly through content and citations, covered in our per engine playbooks.
