We bet on AI and our own solutions

About Equiti. Equiti is a global, regulated broker (forex and CFDs) operating since 2008 — with offices in Amman, Dubai, London and Nairobi among others, a team of more than 170 people and 24/5 customer service. It is a mature brand with real market expertise: its analysts are regularly invited to Bloomberg, CNN and CNBC. Equiti operates natively in English and Arabic, in more than a dozen markets — from the United Kingdom to the Gulf states. In short: a strong foundation of credibility on which SEO makes real sense.

I work as an SEO consultant specialising in semantic search and in content designed for AI models. Most agencies sell hours in other people’s tools: a licence, an export to Excel, the consultant’s intuition. Under Equiti’s visibility I have put something different — my own, connected system in which every decision rests on real data rather than a hunch. These are solutions usually reserved for large product teams: understanding meaning through vectors, parallel AI models, in-house content pipelines and analysis of Google data at scale.

What this gives in practice. We work on the truth from Google Search Console and the real SERP, not on third-party estimates. We scale what others do by hand — thousands of keywords, 100,000+ GSC rows, semantic mapping and linking — without losing quality. And the content is designed from the ground up for the way AI models read and cite it. The result: less guessing, more repeatable, measurable moves that a competitor with five separate subscriptions cannot reproduce.

The goal of the cooperation. To build lasting visibility for Equiti in classic Google and in AI answers — on keywords that can realistically be won, in every language and in every market. This document is the map of that journey: what we are fighting for, how, and with which tools.

Equiti plays in one of the toughest content markets on the web: forex, CFDs, indices, shares, commodities and crypto. Global brands with huge budgets fight for every keyword here, and on top of that come compliance requirements — leveraged products need clear risk warnings in every piece of content. This is not writing a blog “about anything”.

What we are fighting for — four fronts

What we are fighting against — the challenges

• A crowded, global SERP. Forex and CFDs are keywords the whole world races for — you have to pick battles that can be won.
• The era of generative search. More and more answers reach the user from AI before they click a link at all.
• Multilingual content and compliance. EN + AR (RTL), mandatory risk disclaimers — without losing quality or the brand voice.
• Scale and cannibalisation. Hundreds of topics and thousands of keywords where pages can compete with one another — impossible to manage by hand.

When someone asks Google about “what is leverage in forex” or “how to trade gold”, more and more often they do not see ten blue links at the top — just a ready answer generated by AI (AI Overview), stitched together from fragments of a few trusted sources. ChatGPT, Perplexity and Gemini do the same. The effect: being cited in an AI answer is becoming the new “position number 1”, and the user often finishes without a single click.

Modern search stopped “counting words” long ago. Google (thanks to models such as BERT and MUM), and AI engines even more so, understand content through meaning — through entities, intents and the relationships between them. That is why I do not optimise “for a keyword” but for the topic and the user’s intent. I assign keywords to pages using semantic vectors (embeddings), build topical authority and strictly enforce the rule one concept = one page. It is the difference between stuffing keywords and real, complete topic coverage — which both the search engine and AI models reward today.

• Semantics instead of keyword density. We understand intent and meaning rather than counting occurrences of words — the way modern ranking works.
• Data instead of intuition. Every decision is settled by GSC and the real SERP, not by “it seems to me that…”. Where others estimate, we read the source.
• AI-first, not AI bolted on. We design content for model citations from the ground up — because that is where the user’s attention increasingly goes.

The two charts below are an illustrative part of the analysis (sample data) — they show how the audit “sees” the site through embeddings: the site’s topical coherence and the quality of internal linking. These are the two things in SEO that are hardest to show, reduced to two pictures.

And these are not metaphors — this is literally how our audit “sees” the site. The first chart: I turn each page into a meaning vector and project the whole site onto one map; I calculate radius_dist — the distance to the cluster centroid — so “off-topic” is a number, not an impression. The second: for every link I calculate the cosine between the anchor and the target page; below the 0.45 threshold the anchor does not describe the target (generic “here / read more”) and lands on the list to fix. Instead of “improve the linking” you get something specific: this anchor, on this page, does not lead where it promises.

The goals we measure

The cooperation has clearly set, countable goals — not “more traffic some day”, but specific fronts on which we check progress:

A few of them up close — these are live screenshots, not mock-ups:

Cannibalisation is a situation in which several pages of the same site rank for the same keyword. Google does not know which one to show, so it often shows none of them high up — and clicks and impressions are diluted between addresses. You pay for it with traffic that does not show on any “upward” chart.

The mechanism — 5 steps on real GSC data

See it live — one keyword, three competing pages

Keyword analysed: “what is forex” · 248,400 impressions/month (GSC)

Once the competing pages’ content is moved to the target page and redirected with a 301, the signals stop being diluted — one page climbs, takes the impressions of both and moves higher.

This is what the priority report looks like (export from the tool)

The starting point: a raw list of 608 topic proposals. The problem — duplicates, off-topic entries (hundreds of “EUR/USD conversion” pages that do not build a broker’s business) and topics already covered. A raw list is not a plan. We ran it through a five-stage mechanism: currency pair filter → embeddings (recall) → LLM judge (precision) → anti-cannibalisation de-duplication → ranking business core ≫ volume ≫ number of keywords.

Most importantly: each of these challenges already has a mechanism assigned to it. A competitive SERP → ranking by business core. Cannibalisation → the GSC tool with a ready action. Multilingual content → native EN/AR generation. Compliance → disclaimers in the pipeline. Zero-click → content built for AI citations. This is not a list of problems — it is a list of ready levers.