Algorithm Agency

Shortlisted at the 2026 IAB Bookmarks. 🏆Craft : Software, Coding & Tech Innovation. PATHFINDER, our AI demand intelligence platform built in-house, has been recognised by the IAB for the engineering and innovation behind it. #Bookmarks2026 #AlgorithmAgency

We're officially shortlisted for the Mark Awards 2026. 🏆4 entries. 2 categories. Technology + Leadership. PATHFINDER. Agency of the Year. Agency Leader of the Year. AI & Machine Learning Excellence. Proud of everything this reflects. 👀 #MarkAwards2026 #AlgorithmAgency

Question for every CMO: Does your board presentation answer this question? "For every rand we spend on marketing, how much revenue does it produce?" If not, this article is for you. Link : algorithm.agency/blog/cmo-bo…

Your brand appeared in an AI answer. But were you cited OR only mentioned? Citation = AI links to you as the source! Mention = AI names you, links to elsewhere! Same name on the page. Different commercial outcome. Part 3 of the Understanding GEO series: algorithm.agency/blog/citati…

SEO tells you where you rank on Google. It tells you nothing about what happens when someone asks ChatGPT about your category. Those two things are not the same discovery channel anymore, they haven't been for a while. Read our blog that details our AI visibility tool #AI #GEO

Your SEO is up. ROAS is strong. Content calendar is full. And you're only growing at 3%. The problem isn't your channels.n It's the system connecting them. #PerformanceMarketing #Marketing #Digital #ExtremePartnership

Two pools means two work plans, running in parallel. Long-game: feed training data. Multi-year consistent signal across trusted sources. Short-game: win live retrieval. Crawlable, structured, current. Skip either one and you're running a half-strategy. algorithm.agency/blog/where-…

Quick CMO test: Appeared consistently in trusted sources for 18-24 months? → probably in training data. Story changed in the last 6 weeks? → need live retrieval to carry the new version. Memory is still on the old one. Both can be true at once.

Each engine uses the two pools differently: ChatGPT leans more on training data. Perplexity is retrieval-first by design. Claude weighs both. Gemini has direct Google index access. Same query. Four different pool weightings. This is why per-engine GEO strategy matters.

Pool 2: live retrieval. The model's eyes. A page published yesterday can be cited today. Real time. The catch: it gives the model an input, not a belief. The citation can disappear from the next session if a different source ranks higher in that moment. It's less stable

Pool 1: training data. The model's memory. Frozen at the cutoff date. Refreshes every few quarters at best. It rewards consistent signal: a brand mentioned the same way across 10 trusted sources beats a brand mentioned once on a great page.

When AI builds an answer about your brand, it reaches into two completely different pools. One is what it already knows, frozen months ago. One is what it can fetch right now. Most marketers are optimising for one and ignoring the other.

What a CMO should do in the next 90 days: 1. Define your one-line entity 2. Audit your external footprint for contradictions 3. Fix structured data 4. Win third-party signals 5. Measure entity visibility, not page rankings Full guide: algorithm.agency/blog/how-ai…

The SA angle: South Africa's lower web signal density is actually a window. Every new credible mention disproportionately shapes how the model sees your brand here. 83.4% of SA AI citations come from sources that only one engine trusts. The first movers are already compounding.

Three things follow from treating entities as the unit of work: 01. Mentions > rankings. Citations reward your entity, not your page. 02. Consistency > volume. 10 sources saying the same thing beat 50 contradicting each other. 03. Third-party sources do most of the work.

The CMO test: If a journalist asked 5 independent people in your category to describe your brand in one sentence, would they say the same thing? The closer those answers are, the stronger your entity. AI runs that same comparison at scale across millions of sources.

The question has changed. Old: "Is my page indexed?" New: "Does the model know who we are and what we stand for?" Google ranked pages. AI cites entities. A page at position 1 in Google can still be invisible in ChatGPT, Perplexity and Claude for the same query.

AI doesn't index your brand. It builds a profile of it. From every mention, every co-occurrence, across millions of sources. If that profile is fuzzy or absent, you don't get cited. The entity model that decides GEO visibility

We ran 2,737 SA prompts through ChatGPT, Perplexity & Claude. 46,315 domain-citations later: 83.4% of cited domains appeared on only ONE platform. There is no AI search. There is ChatGPT, Perplexity & Claude. They cite completely different SA sources. algorithm.agency/blog/we-mea…