GEO for tech consultancies: get cited by AI

What AI search engines actually cite

Large language models do not rank pages the way Google does. They retrieve and synthesise sources at query time, then cite the ones that read as authoritative, specific and self-contained. The page that gets cited usually has three things: a clear answer to the question in the first 200 words, structured supporting detail, and unambiguous attribution to a named organisation. Marketing pages with vague taglines almost never get cited. Specific guides with named numbers and a clear author do.

The five things to fix this quarter

1. Publish a llms.txt and llms-full.txt that summarises your firm, services and ICP in plain prose.
2. Allow AI crawlers explicitly in robots.txt: GPTBot, ChatGPT-User, Claude-Web, anthropic-ai, PerplexityBot, Google-Extended.
3. Write Organization, Service and FAQPage JSON-LD on every relevant page so models can extract entity facts cleanly.
4. Lead each article with a one-paragraph direct answer before the long-form body. AI engines lift the answer paragraph.
5. Use specific, named numbers (region, sector, currency, time window). Models prefer concrete claims over hedged ones.

Topics that get vendor-aligned consultancies cited

The query patterns that drive AI shortlist citations for consultancies are predictable. They cluster around vendor partner types ('best AWS Premier partners for FSI'), workload questions ('SAP S/4HANA implementation partners in the UK'), comparison queries ('Microsoft Dynamics partner vs in-house build') and ROI questions ('how much does an outbound program for an SAP partner cost'). Each of those is a standalone article opportunity. Generic 'about us' pages are not retrieved.

How to measure GEO without a dedicated tool

There is no Google Search Console for AI engines yet. The pragmatic measurement loop is manual and weekly: define 15 to 25 buyer-intent prompts, run them across ChatGPT, Perplexity, Claude and Google AI Overviews, and log which sources are cited. Track three numbers: presence rate (are you cited at all), share of voice (you vs named competitors) and citation depth (which page is cited). Improvements compound across model retraining cycles, so consistency matters more than volume.

What we do for our clients

GEO is not a separate service for us, it is part of how we build the content and authority layer underneath outbound. The outbound system books meetings now. The GEO layer compounds for the next 18 months and starts producing inbound from buyers who never clicked a Google result. The two channels share the same ICP work, the same positioning and the same artefacts.