Getting cited by ChatGPT and Google AI: GEO in practice

The buying journey has changed. A prospect looking for a provider now asks an assistant directly: 'which accounting firms are there in Tangier for a small business?'. The answer comes with a list and a few links. If your site is not on that list, it did not exist. GEO, or generative engine optimisation, is the methodical response to this change. It is not a new advertising channel, but background work on how information presents itself on the web. Businesses that start early gain an advantage that is not easily caught up: being the usual source on a topic takes months of consistent answers. ## How an AI chooses its sources Answer engines combine several mechanisms: an index of pages, a search at the time of the question, then a synthesis that cites a few sources. Content is more likely to be selected if it clearly answers the question, is dated, is attributed to an identifiable author and is corroborated elsewhere. Three factors weigh particularly heavily. Clarity: a page that answers in a sentence that is visibly extractable is easier to cite than a page that builds its answer over three thousand words. Credibility: a site that publishes sourced figures, an identified team and media mentions inspires more trust. Consistency: when the name, address and service are identical on the website, directories and social profiles, the model has a strong signal. Passages count more than pages. Engines split content into extracts. A paragraph that starts with the question and answers it completely, with no ambiguous pronoun, is far more likely to be repeated than a paragraph that refers back to the previous section. ## Preparing your pages to be cited The work starts with a clarification: which questions do you want to be the source for? A consultancy, a craftsman or an online shop do not aim at the same queries. You need to list the real questions, then answer them precisely on existing pages. Then comes the formatting. You add a short answer at the top of the page, explicit definitions, comparison tables, step lists and figures with their year. Service pages benefit from stating the geographic area actually covered, the prices or ranges and the lead times: this information is often exactly what the user is trying to compare. Structured data help engines understand the nature of the content. These are not keywords: they declare which elements of the page are a question, an answer, an author, a business or a product. They complement already clear content; they do not replace it. A final point on freshness: models give more weight to recent, dated content. A page that shows 'updated in October 2026' and contains figures from that period is more easily cited than an undated page. Updating key pages once or twice a year is often enough; there is no need to publish constantly. ## External credibility No technique replaces presence outside the site. Generative engines favour sources that already exist in several recognised places: local magazines, professional sites, comparison sites, local directories, answers on professional forums. A mention in a local outlet on a relevant topic is often worth more than ten blog articles without citations. And a directory updated with the same name, address and description encourages the retrieval of reliable information. This external presence requires consistency, not volume. Three well-targeted local press articles and a complete business listing produce more than a string of press releases with no follow-up. What matters to a generative engine is the repetition of identical information, attributed to different, credible sources. You also have to accept limited control. A business cannot 'buy' a citation in ChatGPT, and answers change. Monthly tracking of questions and cited sources shows where you are progressing and where content is missing. The full [AI search optimisation (GEO)](/services/seo/geo-referencement-ia/) guide details the method, and the [SEO copywriting](/services/seo/redaction-seo/) guide remains the foundation everything rests on.

When this guide is for you

Your prospects ask questions to ChatGPT or Google AI before searching on a classic engine
You want to be cited as a reference source in your field
Your organic traffic is stagnating even though your classic positions are good
You sell a service where comparing providers now goes through an AI answer

Step by step

Observe what AI already answers — Ask ChatGPT, Perplexity and Google AI about ten real questions from your prospects ('which accounting firm in Tangier for a micro-business?'), note the sources cited and spot those that come up repeatedly. These competitors are your real rivals.
Rewrite pages in 'answer' format — Put a direct two to four sentence answer at the top of each key page, short definitions, lists, tables and dated figures. Engines extract passages: a self-contained paragraph is better than a long introduction.
Make data verifiable — Name the author, the update date, the figures and their sources. Add structured data (Question/Answer, Business, Article). Content that can be cross-checked elsewhere is cited more often.
Build a consistent external presence — Be mentioned in directories, local media, comparison sites and professional profiles, with the same name, address and service information. Models rely on consensus between several sources.
Measure citations — Track every month the questions tested, the sources cited and the clicks coming from AI in analytics. AI traffic is often low in volume but highly qualified.

What it costs (in MAD)

Service	Indicative price	Good to know
GEO audit and tested questions	3,000 – 12,000 MAD (≈ 270 – 1,100 €)	Recording of citations, analysis of cited competitors
Rewriting pages in answer format	800 – 3,000 MAD per page (≈ 70 – 270 €)	Direct answer, lists, structured data
Monthly citation tracking	2,000 – 7,000 MAD per month (≈ 180 – 640 €)	Recurring tests, adjustments, content

Indicative ranges for the Moroccan market in 2026, excluding tax. Prices vary with scope and provider.

Mistakes to avoid

Repeating keywords believing AI engines work like classic engines
Writing pages that contain no explicit answer, only context
Ignoring external sources: a site on its own, with no mentions, is rarely cited
Leaving contradictory information on directories and social media, which blurs the model's trust

Frequently asked questions

Does GEO replace SEO?

No, it complements it. AI engines rely on indexing, reputation and the technical quality of sites. A poorly indexed or slow site is cited less. GEO adds a layer of formatting and credibility.

Can you force an AI to cite your site?

No. There is no equivalent of a bought position in ChatGPT. Models are influenced by quality, clarity and the presence of information on several sources. Some sponsored placements exist in a few products, but they do not replace organic citation.

How long before being cited?

This varies from a few weeks to several months depending on how often models are updated and the competition on the topic. Results are irregular: a site can be cited for one question and not for a close variant.

Should AI robots be allowed to read your site?

It is a strategic choice. Blocking all AI robots protects the content but makes citation impossible. Many companies allow reading and only block use for training, through the robots.txt directives of the relevant robots.

Is traffic from AI measurable?

Partially. Clicks from AI assistants arrive as referral visits that are often poorly identified. What is mainly measured is the change in cited questions, and new contacts are asked how they found the business.

Version française: /insights/etre-cite-par-les-ia-geo-referencement-ia/