Ask Google a question in 2026 and you will often get a written answer before you get a list of links. Ask ChatGPT the same question and you never see the links at all unless you go looking for them. For any business that spent years building rankings, that is a change in how the whole channel pays out.
The numbers are blunt. Pew Research tracked 68,879 real Google searches made by 900 US adults and found that people clicked a traditional search result on 8% of visits when an AI summary was on the page, against 15% of visits when it was not. They clicked a source cited inside the summary on 1% of visits. Separately, Ahrefs analysed 300,000 keywords and put the fall in position one click-through rate at roughly 34.5% once an AI Overview sits above it.
Ranking first still matters. It just buys you less traffic than it used to. The job now has two halves: rank well, and get quoted inside the answer. This guide is about the second half.
AEO, GEO and LLMO are three names for the same job
You will see Answer Engine Optimization, Generative Engine Optimization and LLM Optimization sold as if they are separate disciplines with separate price tags. They are not. All three describe one goal: being the source an AI system pulls from when it writes an answer about your category.
Treat any agency that presents this as a brand new service line with a brand new retainer with some suspicion. Most of the work is technical SEO and content quality you should already be doing. The genuinely new part is smaller and more specific: structuring pages so a model can lift a clean, attributable answer out of them, and building enough independent mentions that the model treats your brand as a known entity rather than a stranger.
What AI engines actually cite
Citation behaviour differs by engine, and that matters for where you spend effort.
- Perplexity cites generously and links prominently. It is the easiest engine to earn visible referral traffic from.
- ChatGPT cites less often per answer but drives the largest share of AI referral traffic by volume, simply because it has the most users. It leans on a small set of high-trust sources, including forums, encyclopaedic references and video.
- Google AI Overviews pull mostly from pages already ranking on page one for the query. If you are invisible in classic organic search, you are invisible here too.
Across all of them, one pattern holds: blog and guide pages get cited more than homepages or service pages. Comparison content, how-to steps, definitions and pages with clearly sourced statistics get picked up far more often than brochure copy. If your website is five service pages and a contact form, there is almost nothing for a model to quote.
The technical layer: make your pages readable by machines
Before content strategy, fix the plumbing. An AI crawler that cannot render your page will not cite it.
- Serve content in HTML, not only in JavaScript. Many AI crawlers do not execute JavaScript the way Googlebot does. If your key copy only appears after a client-side render, it may be invisible. Check by disabling JavaScript and reloading.
- Stop blocking the crawlers you want. Check robots.txt for GPTBot, OAI-SearchBot, PerplexityBot, ClaudeBot and Google-Extended. Plenty of sites blocked these in 2024 and never revisited the decision. Decide deliberately, not by leftover default.
- Add real schema markup. Organization, Article, FAQPage, Product, LocalBusiness and BreadcrumbList. Schema does not force a citation, but it removes ambiguity about what your page is and who published it.
- Fix your headings. One h1, descriptive h2s phrased close to how people ask the question. Models chunk pages by heading. Vague headings produce unusable chunks.
- Keep pages fast and stable. The same speed work that helps rankings helps here, for the same reason: the page has to be retrieved and parsed cleanly.
This is standard technical work, and it overlaps almost entirely with the checklist in our guide on why your website is not ranking on Google. If that audit is already done, you are most of the way there.
The content layer: write pages a model can quote
Models extract passages. They do not read your page for pleasure. Structure accordingly.
Answer first, then explain. Put a direct two or three sentence answer immediately under each h2, then expand. If your answer is buried in paragraph six after a long preamble, the extractable chunk is the preamble.
Be specific enough to be worth quoting. "Affordable web design" is not quotable. "Business website builds in Nairobi typically run KES 120,000 to KES 450,000 depending on page count and custom functionality" is. Numbers, ranges, named steps and dated figures all give a model something concrete to lift.
Cite your own sources. Pages that show where their data came from get cited more often than pages that simply assert. That is not a coincidence. Attribution is a quality signal a model can check.
Write the comparison pages you have been avoiding. "X vs Y", "alternatives to X", "how much does X cost" and "is X worth it" are exactly the query shapes AI answers get built for. Most businesses skip these because they feel awkward. That reluctance is the opportunity.
Date and maintain everything. Show a published and updated date. Refresh figures each year. Stale numbers get quietly dropped from the citation pool.
Off-site: models trust what other people say about you
This is the part most businesses get wrong. AI systems build a picture of your brand from the whole web, not just your website. If nobody outside your domain describes what you do, the model has one source and low confidence in it.
Practical priorities, in order:
- Get listed accurately everywhere that matters in your market. Google Business Profile, industry directories, chambers of commerce and trade bodies. Consistent name, address and description across all of them.
- Earn coverage in publications a model already trusts. One mention in a recognised trade title outperforms twenty links from directories nobody reads.
- Show up where the discussion happens. Forums, Q&A sites and community threads are cited heavily, by ChatGPT in particular. Answer real questions in your category as a named person from a named company. Do not spam.
- Publish original data. A small annual survey of your own customers, a pricing benchmark, a regional trend report. Original numbers get cited and re-cited for years. This is the highest-value tactic on the list, and almost nobody in the mid-market bothers.
How to measure AI visibility when analytics barely shows it
Standard analytics will underreport this badly. Someone reads an AI answer that mentions you, then searches your brand name and arrives via direct or branded organic. The AI touch never appears in the report.
Track four things instead:
- Referral traffic from AI hosts. Build a segment for chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com and claude.ai. It will be a small number. Watch the direction, not the size.
- Branded search volume. If AI visibility is working, people who never heard of you start typing your name. Watch branded impressions in Search Console month over month.
- Manual citation checks. Once a month, run your 15 to 20 most commercially important questions through ChatGPT, Gemini and Perplexity. Log whether you appear, which page was cited, and who was cited instead of you. It is tedious, and it is the most useful data you will get.
- Self-reported attribution. Add a "how did you hear about us" field to your enquiry form. In the UAE and Kenya especially, we now see "ChatGPT" written in by hand on that field.
What it costs and where to start
Because most of this is SEO and content work under a newer label, price it against SEO benchmarks rather than as an extra. Typical 2026 market ranges for a monthly programme that includes AI search work:
- Kenya: KES 80,000 to KES 350,000 per month, depending on content volume and technical scope.
- UAE: AED 6,000 to AED 25,000 per month.
- UK: GBP 1,500 to GBP 6,000 per month.
- US: USD 2,500 to USD 10,000 per month.
A one-off technical and citation audit sits well below those figures and is the right first purchase if you are not yet sure the ongoing spend is justified.
Five questions worth asking any agency pitching you on this:
- Which specific AI engines will you track, and how often?
- Can you show me a citation log for an existing client, with the queries used?
- What are you changing on my site versus off it, and in what ratio?
- How do you handle the fact that AI referral traffic is mostly invisible in analytics?
- What happens to this work if Google changes how AI Overviews source content again?
Any agency that cannot answer the last one honestly is selling certainty it does not have.
Where to put your first 90 days
If you do nothing else, do these in order. Month one: fix crawler access and schema, then audit which of your pages currently rank on page one, since those are your AI Overview candidates. Month two: rewrite your top ten commercial pages to answer first and include specific figures, and add the three comparison or cost pages you have been avoiding. Month three: start the monthly citation log, and commission one piece of original data you can own.
That sequence works because it front-loads the things that also improve classic search. Nothing here is a bet on a single engine surviving in its current form.
Equinode has run search and content programmes for 25+ brands across three continents over 12+ years, and we build this work into strategy, brand, web, content and SEO/SEM as one programme rather than selling it as a separate line item. If you want to see how your brand currently shows up inside AI answers, get in touch and we will run the citation check on your top queries. You can also read our regional deep dive on SEO services in Dubai, or see the full range of what we do on our services page.