To choose an AI search agency, ask how it will measure your visibility in ChatGPT, Gemini, Perplexity, Copilot and Google's AI Overviews and AI Mode, ask for case studies with stated periods and data sources, and walk away from anyone who guarantees citations. The rest of this guide explains what the work involves, the questions worth asking and the warning signs to watch for.
AI search agencies sell the work under several names: AI search optimisation, GEO (generative engine optimisation), AEO (answer engine optimisation), LLM optimisation or AI visibility. The labels differ, but the job is the same: getting your brand named, recommended and cited when people ask an AI assistant instead of scrolling through search results.
What GEO and AEO actually involve
Strip away the jargon and AI search optimisation comes down to four kinds of work. Any agency you hire should be able to explain how it handles each one for your business.
Technical access
AI engines can only use what they can reach. That means checking that AI crawlers such as GPTBot, PerplexityBot and Google's crawlers are not blocked, that important pages load and render without relying on scripts the crawlers cannot run, and that schema markup tells machines what each page is about: your organisation, your services, your locations and your FAQs.
Content written to be quoted
AI answers quote passages, not whole pages. Content that gets cited answers the question in the first sentence, then gives the detail: specific facts, clear headings, short paragraphs and FAQ sections built around the questions people actually ask assistants. An agency should research those conversational questions for your market, not just reuse your old keyword list.
Mentions on other sites
AI engines lean heavily on what other sources say about you. When an assistant recommends providers in a category, it often draws on comparison articles, best-of and top 10 lists, directories, review platforms and industry publications. Good GEO work identifies which of those sources the engines cite for your category, then works to get your brand accurately represented on them.
Measurement
Without tracking, nobody can tell whether the work is helping. AI visibility tracking runs a fixed set of prompts through several AI engines on a schedule and records whether your brand is mentioned, whether your site is cited, which competitors appear and which sources the engines rely on. Combined with AI referral traffic in GA4 and impressions in Google Search Console, it gives you a before-and-after picture.
Questions to ask before you sign
Use these questions on every agency you speak to. The answers will tell you more than any sales deck.
- Which AI engines will you track, and with which tool?Expect named engines, such as ChatGPT, Gemini, Perplexity, Copilot and Google AI Overviews, and a named tracking tool, not a vague promise to monitor AI.
- Which prompts will you track, and who chooses them?The prompt set should reflect how your customers ask about your category, agreed with you before work starts.
- What is my baseline today?A capable agency can show you how often AI engines mention and cite your brand now, and which competitors they name instead.
- What will you do in the first three months?Ask for the specific technical, content and off-site work, in order of priority, rather than a list of service names.
- Who does the work?Find out whether the people writing your content and fixing your site are in-house specialists or outsourced, and who your day-to-day contact is.
- What will the monthly report show?Ask for a sample. It should cover mentions, citations, competitor share and AI referral traffic, and explain what changed and why.
- Who owns the accounts and data?Your Google Search Console, GA4 and any tracking data should sit in accounts you own, so you keep them if you change agencies.
- How does this connect to my SEO?If another agency runs your SEO, ask how the two will work together, and who fixes technical problems that affect both.
Red flags to watch for
Some sales pitches should end the conversation. These are the ones we see most often.
Guaranteed citations or rankings
Nobody controls what ChatGPT, Gemini or AI Overviews say, and the engines change how they answer all the time. An agency that guarantees your brand will be recommended by ChatGPT, or promises a fixed number of citations, is either misunderstanding the channel or misleading you.
No measurement plan
If an agency cannot tell you which engines and prompts it will track, it cannot show you whether its work made a difference. Be cautious of reports built on a few hand-picked screenshots of AI conversations. One good answer on one day proves very little, because AI responses vary from one run to the next.
Vanity metrics
Watch for reports that lead with numbers you cannot connect to the business: total impressions with no period, the number of pieces of content published, or a proprietary score with no explanation of how it is calculated. The numbers that matter are your share of mentions and citations for prompts that lead to sales, AI referral traffic, and the enquiries and sales that follow.
Shortcuts that put your brand at risk
Mass-produced pages, fake reviews and paid placements on low-quality sites can look like quick wins. They can also damage how search engines and AI systems judge your brand. Ask how each tactic would look if a customer or a journalist found it.
AI search sold without SEO groundwork
An agency that sells AI search as something entirely separate from SEO, and never looks at your technical health, page structure or schema markup, is skipping the foundations the engines rely on.
What proof to ask for
Every agency has success stories. Ask for them in a form you can check. A case study worth trusting includes:
- A stated period, such as April to September 2026, compared with a like-for-like period, such as April to September 2025, so seasonality is the same on both sides.
- A named data source: Google Search Console, GA4 or a named AI visibility tool, read from the client's own accounts.
- A clear description of the work: what the agency changed on the site, what content it wrote and what it did off the site.
- A metric that matters: AI referral visits, citations for relevant prompts, non-brand search clicks or enquiries, not only impressions.
- A client you can contact, or at least a named client, where confidentiality allows.
Here is what that looks like in practice. In our ESET case study, visits from AI assistants to ESET's African website rose 134% between April and September 2026, compared with the same months of 2025, measured in GA4 from the client's own account. The claim names the period, the comparison, the metric and the source. Hold every agency, including us, to that standard.
It also helps to ask how the agency would assess its own visibility. AI search agencies compete for the same prompts their clients do, such as requests for the best AI search agency in South Africa. An agency that tracks and improves its own presence in those answers understands the channel from the inside.
What AI search optimisation costs in South Africa
Most South African agencies price AI search inside an SEO retainer or as a monthly programme alongside one, because the technical and content work overlaps. As a typical market range, SEO in South Africa costs R3,000 to R30,000+ per month: R3,000 to R8,000 for small or local businesses, R8,000 to R20,000 for medium businesses, and R20,000 to R50,000+ for eCommerce or national campaigns. Web SEM's own SEO and AI search plans are quoted to fit each business.
What moves the price: how many pages need work, how competitive your category is in AI answers, how many engines and prompts are tracked, and how much off-site work is needed to get your brand onto the sources AI engines cite. Ask every agency to break its fee down into those parts. Web SEM gives a fixed quote after a free consultation.
A simple way to start
Frequently asked questions about choosing an AI search agency
An AI search agency works to get your brand mentioned, recommended and cited in answers from AI engines such as ChatGPT, Gemini, Perplexity, Copilot and Google's AI Overviews and AI Mode.
The work covers technical access for AI crawlers, schema markup, content written to be quoted, mentions on the sites and directories the engines rely on, and monthly tracking of how often the engines name your brand.
Ask for a baseline of your current AI visibility, a sample monthly report and case studies that state the period, the comparison period and the data source.
A good agency names the engines and prompts it will track, explains the work it will do in the first three months and never guarantees citations.
No. No agency controls what ChatGPT or any other AI engine says, so a guarantee of a recommendation or citation is a red flag.
An agency can improve the signals AI engines rely on, such as clear content, schema markup and mentions on trusted sites, and show you the change in your share of mentions over time.
GEO builds on SEO rather than replacing it. Both depend on pages that can be crawled, understood and trusted.
GEO adds work aimed at AI answers: content structured so passages can be quoted, presence on the comparison lists, directories and publications AI engines cite, and tracking of mentions and citations across AI engines.
Most agencies price AI search inside or alongside an SEO retainer. As a typical market range, SEO in South Africa costs R3,000 to R30,000+ per month, depending on business size and competition.
Web SEM quotes every SEO and AI search plan to fit the business, with a fixed fee agreed after a free consultation.
AI search results should be measured with a fixed set of prompts tracked across named AI engines over time, combined with AI referral traffic in GA4 and search data from Google Search Console.
Agree the prompts and the baseline before work starts, and compare like-for-like periods so seasonality does not distort the picture.



