Social Media SEO: What 22 Australian Trade Searches Actually Showed Us

Everyone has an opinion on whether social media helps your rankings. Almost nobody has opened a search result and counted.

So we did. On 19 September 2026 we pulled live Google Australia results for 22 real trade searches, top 20 deep, and recorded every feature, every result and every source the AI Overviews cited. Plumbers, sparkies, concreters, roofers, fencers, plasterers, bobcat hire, pest control. The things a customer types before they ring someone.

Two things came back, and they are both true at once. Social media does not rank your website, exactly as Google has said for over a decade. And social media was sitting in the results anyway, on all 22 searches, with Reddit at position two three separate times.

What We Tested

That last row matters. Checking this in your own browser shows you your personalised Google, not your customer's. And the honest limits: 22 searches, one day, desktop, Australia only. Twelve of the 13 AI Overviews returned a citation list, so that part rests on 98 citations. It is a snapshot, not a law.

Run it yourself. Twenty searches your customers would type, written down with what comes back. It takes an afternoon and you will know your market better than any agency deck will tell you.

What We Found

1. 1Social appeared in the top 20 on all 22 searches. Not always high, but never absent.
2. 2Reddit appeared on 14 of 22 and reached position 2 three times on "roof restoration cost sydney", "is hipages worth it" and "how to find a good tradie australia".
3. 3Facebook appeared on 11 of 22, best position 4.
4. 4YouTube appeared on 11 of 22 but never cracked the top 10. Its best was position 12.
5. 5Instagram appeared on none of the 22. Not once, at any position, on any search.
6. 6Intent decided the screen. Ten searches got an AI Overview and no map pack, seven got a map pack and no AI Overview, three got both and two got neither.
7. 7Ordinary business websites still won, taking 72.9 percent of the top 20 results and 77.6 percent of the AI citations.
8. 8The most cited single domain in the AI Overviews was hipages, quoted 10 times across 98 citations.

Do Social Signals Rank You? Google Says No

Google has denied this consistently for over a decade, and the statements are on the record. Search Engine Journal keeps a dated list of them in its social signals ranking factor rundown: Matt Cutts ruling out follower and like counts in 2014, John Mueller saying social signals do not directly help organic rankings in 2015, Mueller again in 2016 advising people to use social links to add value rather than hoping for rankings, and Gary Illyes in 2017 describing a social link as a single drop in an ocean.

Ahrefs reaches the same conclusion from the other direction in its analysis of social signals, and its reasoning is the useful part: social engagement is cheap to fake, so it would be a terrible thing to rank on, and correlation studies cannot separate it from the hundreds of other variables moving at the same time.

Google does not count your likes. Google does count what your likes cause.

That distinction is the whole thing. A post that gets seen creates what Google does read: somebody searching your business name, somebody leaving a review, somebody linking to you, somebody recommending you in a thread. Social is a cheap machine for manufacturing those. If an agency sells you social signals as a direct ranking factor, ask them for the measurement.

Intent Decides Which Screen You Are Fighting On

The clearest pattern in the data was not about social at all. It was about which Google the customer gets, and the trigger was what they typed.

The split in one picture. The table above shows a sample of which searches landed where.

Ten of the 22 showed an AI Overview with no map. Seven showed a map with no AI Overview. Three showed both, and two showed neither. So the split is strong but it is not a wall, and the three that showed both are worth noting: "best plumber perth", "excavator hire perth" and "asbestos removal cost perth" are all hiring queries that Google now also answers with a summary. If that keeps spreading, the map pack stops being the whole of local search.

Practically, you are fighting on two screens. Your Google Business Profile wins the hiring screen: reviews, recency, response rate. Content wins the research screen: a page that answers the exact question with real numbers.

Most trade businesses put everything into the hiring screen, because it pays this week. The research screen is where the customer decides what a fair price is and who sounds like they know their trade, and they do that before they ever type your suburb.

Reddit, Facebook And YouTube: The Numbers

• On "roof restoration cost sydney", Reddit sat at position 2, above every roofer in Sydney.
• On "is hipages worth it", the top of the page was Reddit at 2 and Facebook at 4. People asking whether a lead platform is worth the money read each other, not the platform.
• On "best plumber perth", three separate Reddit threads appeared in the top 20.
• YouTube was the volume player and the weak performer: 45 results across 11 searches, none of them in the top 10.

Instagram deserves a blunt note. Zero out of 22 is not a strategy for search visibility. It may still earn you work through the feed itself, but it is not putting you in front of somebody who typed your trade into Google. If you are posting there out of obligation and getting nothing back, put that hour into Facebook, YouTube or your profile.

Who The AI Overviews Actually Cited

Thirteen of the 22 searches returned an AI Overview, and 12 of those gave us a citation list. That is 98 citations. Ordinary business websites took 76 of them, 77.6 percent. Directories and lead platforms took 15, 15.3 percent. Social took 7, 7.1 percent, all of it Reddit and YouTube.

The single most quoted domain was hipages, cited 10 times. Worth sitting with if you pay them for leads: on cost and "is it worth it" questions, the platform you buy leads from is the source Google reads out to your customer.

On "concrete driveway cost perth" the overview cited seven sources and every one was a Perth concreting business. Not one social result, not one directory.

This lines up with the largest independent measurement we could find. Otterly's 2026 AI citation study looked at more than 100 million citations across ChatGPT, Google AI Overviews, Google AI Mode, Perplexity, Copilot and Gemini, and found social and video platforms took 5.54 percent of all citations, with brand domains on 52.2 percent. Our 7.1 percent social on Australian trade queries is the same shape.

It is worth knowing why the headlines read the other way. You will see it reported everywhere that Reddit and YouTube are the most cited sources in AI search, and inside the social slice that is exactly right: Otterly has Reddit at 46.4 percent of social citations and YouTube at 31.8 percent. But the slice itself is 5.54 percent of everything. Reddit dominating social is not the same as Reddit dominating AI search, and a tradie reading the headline could easily conclude their own website does not matter. Our numbers say the opposite.

So if you want to be in the AI answer for what a driveway costs in Perth, publish what a driveway costs in Perth, on your own site, with real numbers. That is the same mechanism we set out in our guide to generative engine optimisation versus SEO: AI answers are assembled from retrieved pages, and the more sources that agree about you, the more quotable you are.

Nobody Is Searching For "Social Signals"

While we had the data open we checked demand. Australian searches per month:

Ten a month for the phrase the whole industry argues about. Meanwhile "create a facebook business page" gets 1,300 a month in Australia on its own, 130 times the demand for the theory.

Australians are not searching for the argument. They are searching for the doing. Write for the second group.

The 60 Day Test, Designed Properly

Nothing above tells you what social consistency is worth in your business. Only a test does, and only if it is designed so the answer cannot be argued with afterwards.

1. 1Pick 5 properties to treat and 5 to leave alone, matched on trade, city and current lead volume. Without the control you will credit social for a seasonal bump.
2. 2Write down the baseline before you start. Calls, form fills, profile calls, direction requests, messages, review count and the date of the last review. Screenshot it.
3. 3Run one treatment. Two to four posts a week, three to five local groups, a reply to every comment and message within 24 hours, a weekly review request, and a link to the profile in the posts. Do not also rebuild the website or you will never know what worked.
4. 4Tag every link so social traffic is attributable rather than a story you tell yourself.
5. 5Run it 60 days, then compare the treated five against the control five. Not against last month, which had different weather and a different economy.

Decide before you start what result would make you stop, and do not change the treatment in week three because it looked flat. This is the design we are putting into our own properties, and we will publish what comes back on this page, including the boring result where nothing moved. If you run it too, send us your numbers and we will publish those alongside ours.

What To Do In The Next 30 Days

1. 1Search your own trade and city twenty times and write down what you see. Which searches show the map, which show an AI answer, and whether a Reddit thread is beating you. Free, and the most valuable hour on this list.
2. 2Claim and clean up your Facebook page. Correct name, suburb, phone, services, linked to the site and the profile.
3. 3Turn on messaging and route it to your phone. An unanswered channel is worse than no channel.
4. 4Answer the price question on your own website. The one every customer asks on every quote, with real numbers and what changes them. This is your AI Overview play.
5. 5Ask for a review at the end of every job for 30 days, and reply to all of them.
6. 6Join three local groups and answer questions. No links, no pitch.

None of it is clever. All of it gets skipped because it is not exciting, and a year later someone with half your skill is outranking you because they did the boring version consistently.

Frequently Asked Questions

If you want a hand designing the test for your own business, have a look at how we build lead generating sites for trade businesses, or work out what one job is actually worth to you with the lead value calculator first.