AI search optimisation in-house or with an agency? How to choose the right operating model

A business that decides to invest in AI visibility usually asks the next question straight away: who should do the work? The answer does not depend on the size of the business or its budget, but on what is missing from its current activities – knowledge, time or coordination. Each of these gaps calls for a different solution. That is why choosing the right approach starts with assessing your own resources, not with comparing proposals.

We most often see businesses that already work with an SEO agency, yet nobody monitors their AI visibility. This is rarely the agency’s fault. The agency delivers SEO under a contract for Google rankings, and the question of how AI assistants describe the brand never makes it onto anyone’s task list. So the first question is usually “whose remit is this?” rather than “who would do it better?”.

This article draws on Google and Bing documentation, research from 2025–2026 and our own experience of AI search optimisation projects.

The article in brief

An SEO team can handle many AI search optimisation tasks on its own because, according to Google, optimising for the AI features in its search engine is still ordinary SEO. External support makes sense where a specific skill, time for regular work or a process owner is missing. Before you hire another supplier, check what your current team or agency can do for you.

Key takeaways for businesses:

1. The best operating model follows from the type of constraint, not from the size of the business. A lack of knowledge, a lack of time and a lack of coordination need different solutions, even in businesses of a similar size.
2. Skills and time are two different resources. A highly knowledgeable specialist who has no time for critical tasks will not deliver the plan.
3. AI search optimisation builds on SEO, but it is not the same thing. It adds, among other things, PR work and regular monitoring of AI model answers, which SEO teams often do not cover.
4. First, check whether your current SEO is built around customer intent. If it is, extending your existing activities with elements of AI search optimisation is usually enough.
5. Doing it yourself is not free, and an agency does not take the work off your hands. On one side, there is your team’s time and learning; on the other, your expertise, approvals and implementation.
6. With two agencies, it is hard to attribute results, and every model needs a single owner within the business. SEO and AI work on the same content and the same website, so someone has to hold the plan together and be accountable for the whole.
7. The model can change over time. A business can start with an audit or monitoring with a partner and then take over the work, or gradually widen the scope of support.

Neither in-house work nor an agency can guarantee mentions, citations or recommendations in AI answers. They can only increase the chances of them appearing. A well-chosen model, however, lets you work consistently, measure results and make the necessary changes. We explain how the whole process works in our guide to AI search optimisation.

Can you do AI search optimisation in-house?

Yes. A business can keep the information on its website accurate and clear, present its offer properly, develop expert content, check selected customer questions in AI assistants and use dedicated tools. An experienced SEO team can also run an audit and look after the technical side of the website. So the line is not drawn between the business and the agency, but between one-off tasks and a regular process.

Google states plainly that to appear in AI Overviews and AI Mode, there are no additional requirements beyond SEO fundamentals. In its July 2026 guide, it adds that optimising for generative search is still SEO, with no new files for AI and no need to split content into small chunks. Both statements, however, apply to Google Search, not to ChatGPT, Claude or Perplexity, which have their own mechanisms for searching and selecting sources. For an SEO team, this means it does not start from scratch with AI Overviews and AI Mode.

The team has basic measurement at no extra cost: Google points to the “Generative AI performance” report in Search Console, and since February 2026, Bing Webmaster Tools has shown a site’s citations in Microsoft Copilot and Bing’s AI summaries, albeit without click data.

The same tasks look different as a one-off activity and as an ongoing process. The table below shows which scope an in-house team usually handles and which needs a specialist or more time.

Task	Usually within an in-house team’s reach	Usually needs a specialist or more time
Up-to-date information about the business and its offer	edits on the website	consistent data across many sites
Expert content	developing existing pages	an ongoing content plan covering many topics
Checking AI answers	a few customer questions	regular monitoring of many questions and competitors
Search Console and Bing reports	periodic data reviews	a combined report from multiple sources
Technical audit	a smaller website	a large site with many templates
Mentions in external sources	business profiles, industry directories, partner websites	mentions and links in industry media

Your team can check an AI answer on its own, but regular monitoring is separate work. AI answers vary, so you only get a reliable picture from many questions checked in a repeatable way. According to research by SparkToro and Gumshoe.ai from late November and early December 2025, the chance that ChatGPT or Google’s AI gives the same list of brands in two answers is less than 1 in 100. The study was not peer-reviewed, and Gumshoe sells a monitoring tool, but the result shows well that a single check gives only a snapshot.

One-off tasks do not yet make a process, so the most important question is: who will carry them out regularly?

How do you check whether your team has the skills and time for AI search optimisation?

Assess six areas of work and ask three questions for each: does someone know how to do it, do they have time for it and can they implement it? The answers will show what your business lacks: knowledge, time or coordination. Each of these three constraints calls for a different solution, so the diagnosis should come before choosing a supplier.

An SEO team already has many of the skills, because in Google Search it works on the same foundations; the gaps usually appear away from the website. In an Ahrefs study from May 2025 covering 75,000 brands, brand mentions correlated more strongly with visibility in AI Overviews than links did: 0.664 compared with 0.218. This is correlation, not proof of causation, but it shows that a brand’s presence in external sources is a separate area of work.

Similarly, in Muck Rack data from May 2026, earned media accounted for 84% of citations in ChatGPT, Claude and Gemini. This is research by a PR tools provider, using a broad definition and data mainly from English-speaking markets, so do not apply the figure directly to your own market. It does, however, point to a skill that SEO teams often lack: PR and building a brand’s presence in industry media.

The second gap concerns how work is organised. In a study by Corporate Ink and Prodedge of 150 B2B technology marketers in the US, around half cited limited knowledge of GEO, and 39% a lack of coordination between content, PR and website teams. Despite the small sample, the result suggests that the problem is often not the tools, but the absence of someone to bring several teams’ work together.

The checklist below will help you assess this.

Area	Does someone know how?	Do they have time?	Can they implement it?
Visibility analysis	Can they select customer questions?	Will they analyse competitors too?	Will the findings make it into the work plan?
Technical	Do they know technical SEO and structured data?	Do developers have capacity for fixes?	Will changes go through without lengthy approvals?
Expert content	Can they write content that AI can cite?	Will experts find time for interviews?	Can they publish and update content?
Credibility and mentions in external sources	Do they have PR experience?	Will they work on it regularly?	Can the business speak publicly?
Monitoring AI answers	Can they tell a mention from a citation and a recommendation?	Will they repeat the checks?	Do they have access to tools and data?
Coordination	Can they put together a joint SEO, content and PR plan?	Will they manage it on an ongoing basis?	Can they set priorities across teams?

If the knowledge is there but time is short, you need delivery support. If knowledge is missing in one area, targeted expert support is usually enough. If nobody can implement changes or set priorities, you lack a process owner, and that requires changing how work is organised, whichever model you choose.

So not every business needs a dedicated GEO specialist, because the checklist often reveals a gap in just one or two areas.

Which delivery models for AI search optimisation can a business choose from?

A business can choose from at least five delivery models for AI search optimisation: an in-house team, extending the work of its current SEO agency, a specialist or freelancer, an agency handling both SEO and AI, and a hybrid model. They differ in who brings the knowledge, who has time to do the work and how much work remains with the business.

Model	When it makes sense	Advantages	Limitations	Business involvement
In-house team	The team has SEO skills and time for regular work	Knowledge of the offer; skills stay in the business	PR and monitoring are often missing; day-to-day tasks crowd out AI search optimisation	Highest: the business plans, delivers and measures
Extending the current SEO agency’s work	SEO has sound foundations, and the agency has time for a wider scope	Continuity, familiarity with the website, one plan	May lack skills in building mentions and monitoring AI answers	Medium: expertise, approvals, some implementation
Specialist or freelancer	One skill is missing	Targeted support, flexible scope	Limited delivery capacity, reliance on one person	High: the business coordinates and does the rest
Agency handling SEO and AI	Time and several skills are missing at once	One supplier is accountable for results in both areas	Knowledge accumulates with the partner; changing partners requires handing over documentation	Lower, but not zero: knowledge of the offer, approvals
Hybrid model	The team covers some areas, and a partner fills in the rest	The business keeps the direction; the partner adds skills and time	Requires a clear division of tasks and an owner within the business	Medium: the business does its part and coordinates

A variant with two agencies (SEO and AI) is also possible, but it requires a careful division of responsibilities, which we discuss in the section on dividing responsibilities.

The partner’s scope can also be scaled. The narrowest scope is monitoring AI answers and setting a direction, which the team implements itself. A wider one includes support with content writing and building mentions and links in credible external sources. The widest is taking over everything, although knowledge of the offer and approvals stay with the business.

Can your current SEO agency handle AI visibility?

Often, yes – provided the SEO has sound foundations and is built around customers and their intent, rather than keywords chosen by search volume. Our audits show that with this approach, the team already knows how to address customer intent, so it is enough to extend current activities more boldly and add monitoring of AI answers.

If your agency has not worked on AI visibility so far, it was usually simply not in the contract. Before you start looking for a new partner, ask them four questions:

1. Is AI visibility within the scope of our contract?
2. Do you monitor AI answers and distinguish a mention from a citation and a recommendation?
3. Do you build the brand’s presence in external sources?
4. Do you implement changes, or only recommend them?

A “GEO” label in an agency’s offer does not confirm these skills, so check your current agency first and only then decide whether to replace or supplement it.

What does in-house AI search optimisation really cost compared with working with an agency?

The cost depends on the scope of work, so compare the total cost for the same scope, not an agency invoice against zero. In-house delivery costs team time, learning, tools, coordination and tasks postponed until later. Working with an external partner, on the other hand, does not relieve the business of providing expertise, approvals and implementation on its side.

The table below shows what makes up the cost in both options and which items stay with the business even when working with a partner.

Cost component	In-house team	External partner
Working time	Team time taken away from other tasks	The partner’s fee plus team time spent on the collaboration
Learning and skills development	Learning and testing, especially in PR and monitoring	On the partner’s side; the team still needs to understand the recommendations
Monitoring tools	Licences and time for regular question checks	Usually within the partner’s scope; agree the business’s access to the data
Coordination and meetings	Setting priorities across SEO, content, PR and IT	Meetings, reports and passing information to the partner
Approvals and expertise	On the business’s side	Still on the business’s side
Technical implementation	The business’s IT team or developer	The partner or the business’s IT team, depending on the contract
Opportunity cost	Tasks postponed while the team works on AI	With two agencies: coordination between them and disputes over attributing results

Start by describing the scope: which areas are to be covered, how regularly and who is responsible for implementation. Only then price both options, including the time that still stays with your team when working with a partner. Comparing invoices alone, without the scope of responsibility, leads to the wrong conclusions, because a cheaper proposal may simply shift more work onto the business.

The cost is determined by how work is divided between the business and the supplier, not by the form of collaboration itself.

How should responsibilities be divided between the business and an external partner?

For each area of work, agree who prepares the activities, who implements them and who approves them. Knowledge of the offer, customers and experience always comes from the business, whichever model you choose. Remember, too, that a partner preparing a recommendation is not the same as implementing it.

The table below shows an example division when working with an external partner. It is one possible arrangement, not the only right one – in a hybrid model, for instance, the business’s team will take over part of the implementation.

Area	Prepares	Implements	Approves
Strategy and priorities	partner (visibility analysis)	partner and the business’s team	process owner within the business
Expertise and information about the offer	experts and sales within the business	partner turns it into content	the business
Content	partner or the business’s content team	person with access to the website	subject-matter expert within the business
Technical changes to the website	partner (recommendations)	IT department or website developer	process owner within the business
Mentions and presence in external sources	partner	partner with the business’s PR person	the business
Monitoring and reporting	partner	partner	process owner within the business
Overall coordination	process owner within the business	process owner within the business	board or marketing director

The most important row in the table is the last one. As we showed when assessing skills, the problem is often not the tools, but the absence of someone who brings content, PR and website work together into one plan. So appoint one person within the business who sets priorities, gathers approvals and makes sure recommendations get implemented.

In our experience, splitting the work between two agencies, one for SEO and one for AI, is possible, but it creates two problems. The two suppliers’ skills overlap, because they work on the same content, website and mentions. It is also hard to establish who is accountable for results, because work done with AI in mind also improves visibility in traditional search. That is why we believe it is best when SEO and AI are not split between two suppliers, but run within a single plan. If a business keeps two agencies, it needs one plan owner and a shared report for both sides.

A responsibility table only works when a specific person is in charge of coordination, not the whole marketing department.

How do you choose the right AI search optimisation model for your business?

Start with the type of constraint, not with a list of suppliers. Use the checklist to establish which gap applies to your business, and remember that the same team may need an expert in one area and only extra delivery capacity in another. The three scenarios below are illustrative and differ in the type of gap, not in the size of the business.

SEO is built around customer intent, and the team or agency has time

This scenario applies to businesses whose SEO has sound foundations and answers customers’ questions, not just high-volume keywords. In our experience, such a team or agency already knows how to address customer intent, so a new supplier is usually not needed. Extend current activities to cover more customer questions and add monitoring of AI answers. The process owner, expertise and content approval stay within the business. The recommendation would change if the team ran out of time or if it turned out that the brand barely appears in external sources.

One skill is missing

The team handles most of the work, but the checklist shows a knowledge gap in one area. This might be, for example, building the brand’s presence in industry media and websites, or regular monitoring of many questions and competitors. In that case, choose targeted support: a specialist or freelancer, or a partner who takes over just that one area. Content, technical work and coordination stay within the business, and the external supplier reports to the same process owner. The recommendation would change if the knowledge gap covered several areas at once or if the team had no time to implement the specialist’s recommendations.

Time and coordination are missing, or SEO targets keywords rather than customers

This scenario covers two situations. In the first, the team knows what to do but lacks delivery capacity or someone to connect content, the website and work in external sources. In the second, SEO is based on high-volume keywords, so the approach needs to change first. In both cases, consider an agency handling SEO and AI or a hybrid model, in which the partner does part of the work and the business’s team keeps the direction. One process owner, knowledge of the offer and approvals stay within the business. The recommendation would change if the business decided to expand its own team, which would take over delivery over time.

The model you choose is not a permanent decision. You can start with monitoring and setting a direction alone, and widen the scope of collaboration when the team cannot keep up. You can also start with an audit with a partner and then hand the work over to your own team once it has gained the necessary skills.

A good model addresses today’s gap and can be changed once that gap disappears.

What mistakes should you avoid when choosing an operating model?

The most common mistakes in choosing a model stem from assumptions made before the business checks its own resources. It assumes that in-house work costs nothing, that an agency will take all the effort off its hands, or that a new label in an offer means new skills. Each of the myths below leads to a model that does not match the real gap.

• “In-house means free.” An in-house team pays with its time, learning, tools and postponed tasks. Compare both options for the same scope of work.
• “The agency will take on full responsibility.” A partner does not know the offer and customers as well as the business does, and someone has to approve and implement its recommendations.
• “We need a dedicated GEO specialist.” A new role only solves the problem when knowledge is missing in several areas at once. Before you hire someone new, check which area really lacks knowledge.
• “An agency with GEO in its offer knows AI.” A new service name is sometimes old SEO in new packaging. Skills are shown by how AI answers are measured and the scope of implementation, not by the name of the service.
• “An AI file or bought mentions will do.” According to Google, its search engine does not require new files for AI (e.g. llms.txt), and chasing inauthentic mentions achieves little.

These mistakes have one thing in common: the decision about the model is made before diagnosing what the business really lacks.

Summary: diagnose your resources first, then choose the model

So the question is not whether to do AI search optimisation in-house or with an agency, but what your business lacks today: skills, people to do the work or someone to hold everything together. The answer determines whether it is enough to extend current activities, bring in targeted expert support, or hand a wider scope to an agency handling SEO and AI or opt for a hybrid model.

The most common starting point is the SEO team or agency the business already has. If their SEO is built around customer intent, the foundations are already in place and mainly need to be developed more boldly. One thing remains constant. Whatever the model, someone within the business should own the plan and the knowledge of the offer.

Start with the checklist in this article and a conversation with the person or agency that runs your SEO today. Establish what is really missing, and only then decide what support you need.