AI Weekly: AI Search Spreads to Social — June 15-21, 2026

AI search jumped from Google to social media this week. Meta added an AI Mode to Facebook that answers questions using public posts across its platforms. At the same time, two more open-source coding models landed, keeping the pressure on the paid options. The through-line: the way people find businesses is changing, and the tools to build with keep getting cheaper.

The Deep Dive: AI Search Spreads, Open Models Keep Coming

Meta brings AI search to Facebook

On June 15, Meta launched AI Mode on Facebook, a way to search that uses Meta AI to surface answers pulled from public posts across its platforms. This is the same shift we saw with Google, now happening on social. People increasingly get an AI-summarised answer instead of scrolling a feed or a list of pages.

For any business that relies on Facebook for reach, this matters. If customers get their answer from an AI summary, your organic posts have to be clear and useful enough to be the thing that summary draws on. Vague, salesy posts do not get quoted; specific, helpful ones do.

More open coding models arrive

The open-source run continued, with Z.ai releasing GLM-5.2, a large model with a very long context window, and Moonshot AI shipping Kimi K2.7 Code, aimed at coding and agent work. Neither is a name your customers know, but each gives developers a cheaper, self-hostable option and keeps the paid providers honest on price. That steady competition is a big reason the cost of building with AI has kept falling all year.

Why this matters:

• AI search on Facebook changes how customers discover you, the same way it did on Google.
• Clear, genuinely useful social content is what AI summaries pull from, so quality beats volume.
• A steady flow of open coding models keeps making custom software cheaper for small teams.

What This Means for Australian SMEs

If Facebook is part of how you get found, treat AI Mode the way you should treat Google's AI search. The goal is to be the useful, factual source an AI wants to quote. That means posts that actually answer a customer's question, not just promote a sale. Think "here is how to choose the right X" rather than "buy our X now".

The open-model news is more of a background win. You will probably never run these models yourself, but they keep downward pressure on what you pay for AI, whether directly or through the tools your suppliers use. Cheaper inputs for them tends to mean better value for you.

The practical takeaway for this week:

• Audit your recent Facebook posts. Would an AI trying to help a customer actually quote any of them?
• Shift some social content toward answering real questions, not just pushing offers.
• When commissioning software, ask whether an open model could do the job for less.

Helping Sydney businesses stay findable as AI search spreads across Google and social is exactly what we do at IntelliAgent. If you want a review of how discoverable your business is, get in touch.