An AI SEO agency for UK tech companies.

What is an AI SEO agency?

An AI SEO agency works to get your company named, and recommended, in the answers AI gives, not only ranked in Google's links. ChatGPT, Gemini and Perplexity each name only a few companies when a buyer asks, so the work is getting you onto that shortlist.

Avouch is a UK AI SEO agency for tech companies.

Specialists call it generative engine optimisation (GEO). There’s a difference between showing up and being recommended: an engine can name you and still send the buyer somewhere else. So we measure your Recommendation Rate™, how often AI picks you as the answer.

We test up to 9 engines: ChatGPT, Claude, Gemini, Perplexity, Grok, Google AI Overviews, Google AI Mode, Brave and Copilot. Each one can give a different answer to the same question, so we read each one on its own.

What is the difference between AI SEO and ordinary SEO?

Ordinary SEO gets you a high place in Google’s links. AI SEO gets you named, and recommended, in the answer an AI engine writes. Half of this is ordinary SEO. Here is which half.

SEO and the AI work are one job, not a separate invoice. AI engines often search the web before they answer, and a page that ranks is easier for them to find, so the ordinary half matters.

What should an AI SEO agency do?

1. 01

   Start from what your buyers actually ask

   Before it writes a word, it should read what buyers in your category post on Reddit, G2 and review sites. We often read thousands of posts, and use them, with what people search for on Google, to choose the questions we put to up to 9 AI engines.
2. 02

   Show you who AI shortlists instead of you

   It should show you which competitors AI names over you, and why. Usually it isn't the biggest one.
3. 03

   Be judged on your pipeline

   It should track how often AI picks you, every week, and let you judge it on pipeline, not on the score. We track your Recommendation Rate™ every week.

What Avouch built to measure AI answers, and what you get from it

We built our own system to test what AI says about you and your competitors, and to write from facts we checked. Here is what you get from it.

• You see whether you are moving. We put a fixed set of buyer questions to AI engines on a schedule and keep the answers word for word, with the engine and the date. Not one screenshot taken on one afternoon.

• You see who AI actually puts forward. Not just whose website it links to. Being linked and being recommended are different things, so we count them separately, for you and for each competitor on your list.

• You see where Google and AI part company. For every question we track, we also keep the ordinary Google results page from the same day, so you can see where ranking well and being recommended split.

• You see what Gemini goes looking for. When Gemini answers one of your buyers' questions, it can run searches of its own first. We record those searches and check whether your site has a page on each topic.

• Our AI writer works from facts it read on the day it writes. It builds each piece from a sheet of facts it read that day, each with its source and date, and checks the draft against that sheet. A piece that fails its checks is held back, not handed over.

• We do it on our own site first. Every page here also exists as plain text for AI tools, with the same words, not a separate version for bots. Add .md to the end of this page’s address to see it. Our page for AI assistants says plainly what Avouch is and what it does not do.

Is there an AI that can do SEO?

Yes, AI can do a lot. We use it ourselves, to read what buyers post and to judge which company, if any, each answer picks. But it is human judgement that sits on top: what to measure, what moves the needle most, where AI adds value and where it causes complications.

That judgement is also what builds the right AI systems, and keeps what they make from being slop. Deciding which page to write, and which site to be on, is the judgement you pay an agency for.

How do you know AI SEO is working?

The unit is your Recommendation Rate™: how often AI picks you when your buyers ask. We report it engine by engine, with the questions and the answers behind it.

Then we look at what it did to your enquiries and your pipeline, and at the other things that move with it:

• What buyers tell you. Sales self-attribution: new buyers saying they found you through AI.

• Direct traffic. An increase in people coming straight to your site, where that is relevant.

• Referral traffic from AI. Visits that arrive from ChatGPT, Gemini, Perplexity and the rest.

• Google and Bing. The AI insights in Google Search Console and in Bing Webmaster Tools.

• Citations and recommendations. How often AI cites your pages, and how often it recommends you.

How much does AI SEO cost?

The free scan costs nothing. The assessment is paid once, and is not part of the retainer. Then there is a retainer.

• The free scan. A small snapshot of the assessment. It costs nothing.

• The assessment. Where you stand across up to 9 engines, and why. Paid once, not part of the retainer.

• The retainer. We do the work.

What do bad AI SEO agencies do?

• They show you answers from their own ChatGPT. Their account already knows your name and your website, so the answer flatters you. We ask from clean sessions that know nothing about you.

• They ask questions that already name you. Ask AI whether your company is a good choice and it will usually name you. That makes you look bigger than you are. We ask the questions your buyers ask.

• They don't know how to set up a site for AI. AI tools blocked from the site, words hidden behind code the tools can't read, page data that doesn't match the page. We fix the site first.

• They test once, on one afternoon. AI answers change from one run to the next, so one screenshot proves nothing. We ask the same questions on a schedule.

• They give you one blended score. Each engine answers differently, and one number hides it. We show you each engine.

• They count a mention or a link as a win. Being named, or linked, is not being recommended. We count recommendations.

• They promise results by a date. “Cited in 30 days” is a promise nobody can keep. We measure where you stand first.

• They fill your site with AI slop. Generic pages that say what every competitor says give AI no reason to pick you, and made-up facts can hurt you. We write only from facts we checked.

• They sell llms.txt as the fix. It is a small text file for AI tools. We have one too, but on its own it does not get you recommended.

• They buy or fake mentions. Planted Reddit posts, fake reviews, paid “best of” lists. They can backfire, and we don't do them.

Which AI SEO agency is the best?

The one that can show its working, and Avouch can. We show you, engine by engine, the questions we put to AI, the answers that came back and who was named instead of you. We can do that because we built our own measurement in-house, and we publish our own research on what makes AI pick a company.

Put these five questions to any agency. Our answer sits under each one.

1. Which buyer questions did you test, and where did they come from?From what buyers write on Reddit, review sites and forums, from what people search for on Google, and from variants of those questions that AI writes for us.
2. What did each engine say on its own?You get a separate figure for each engine.
3. Was I recommended, or only mentioned?Both, counted apart, so you can see the gap between them.
4. Where is your own research, with the method and the dates?In Avouch research.
5. What will you not do?We do not build links, and we never buy or fake citations or placements.

What research is Avouch's AI SEO work based on?

Our work is based on three things: our own research, the results of the clients we learn from, and the best research the sector puts out. Our models are updated every week to account for the changes we see.

For our own research, we put buyer questions to AI engines, read which source made each answer pick a company, and publish what we find. Being cited is not the same as being the reason AI picks a company.

In August 2026 we read 6,517 AI answers from two service markets, engine by engine. On Google AI Overviews, YouTube was cited in 1,216 of them and decided none of the picks.

The full study is What actually decides who AI recommends. Every study, with its method and dates, is in Avouch research.