What we build

What SEO becomes when the answer is written for you

Search used to end with a list of links. Increasingly it ends with an answer, and the only question that matters commercially is whether your business is named inside it.

Three problems that get treated as one

AI SEO services are now three separate jobs. Ranking in classic search results, being quoted in the answer box above them, and being named when someone asks ChatGPT or Gemini about your category. Each has its own fix and its own way of being measured. The first useful step is finding out which of the three you are currently losing.

Where this has actually run. Across client accounts Wobble operates, published reach includes 3.97 million lifetime impressions for Palm Berries and more than 987,000 for Center for Sight. See the work, with the numbers.

SEO is the old problem: appearing in a ranked list of links. It still matters, because the pages that rank are also the pages the answer engines read.

AEO is being the text that gets extracted into the answer box, which rewards a page that answers a specific question in a specific place rather than one that ranks well on average.

GEO is being named when someone asks an assistant for a recommendation. That is not a ranking problem at all. It is an entity problem, and it is won mostly off your own website.

Why GEO is not SEO with new letters

When an assistant recommends companies in a category, it draws on what it has read across the web about who exists in that category. Your own website is one voice among many, and it is the one the model trusts least, because every company describes itself favourably.

Which means the work that moves generative visibility is mostly not on-page. It is being mentioned, listed, discussed and reviewed in places the model already reads. A business with a technically perfect website and no external footprint will be invisible to this entire class of query, and no amount of on-page work will fix it.

A model recommending a company is repeating a consensus it read somewhere. If nowhere says your name, nothing on your own site changes that.

The entity problem underneath all of it

Before any of this works, a machine has to be able to answer one question with confidence: what is this business, exactly. That sounds trivial until you audit a site and find the company described differently on every page, under two different identifiers, with a service area that contradicts its own sales pitch.

Fixing that is unglamorous and it comes first. One name, one description, one identity, corroborated by profiles that resolve to the same organisation. It is the difference between a machine treating your mentions as one company and treating them as several.

What the work looks like in practice

Technical foundations first, because none of the rest is measurable while a crawler is being redirected or a page is claiming two addresses for itself. Then entity consistency. Then content that answers real questions rather than covering topics. Then the off-site work that generative visibility actually depends on.

Measurement runs alongside all of it. Rankings alone stopped being sufficient the moment answers started replacing links, so the tracking has to include whether the business is being named in generated answers, on which queries, and against whom.

How this is scoped and paid for

Visibility work is ongoing by nature and honest providers say so. There is usually a fixed first phase, the audit and the corrections that follow from it, and that phase has an end. What comes after is monthly, because the measurement only means anything if somebody takes it repeatedly and acts on what moved.

That leaves two ways to buy it rather than the four the rest of the systems offer. Either the search system is built and your team trained to run it, a one time build fee with a small retainer afterwards, or Wobble runs it and you rent the output as an agreed monthly list of search work. Adverts, content and search are grouped together on this point for the same reason, and the four ways to own this sets out why they sit outside the grid.

What moves the figure here: how many pages already exist and what state they are in, since correcting a hundred pages is a different job from writing ten; how many markets or languages you need to be found in; and how fast copy can be reviewed and published on your side, because a review queue is the most common reason this work stalls and stalled work is the most expensive kind.

Before quoting, anyone should have measured where you currently stand. Not estimated it. A recorded set of queries, run, with the answers saved, so that in three months there is something to compare against. A proposal with no baseline in it is a proposal that can never be shown to have failed.

When search is the wrong lever entirely

If a business needs revenue this quarter, search is not the lever. Search compounds and it compounds slowly, and anyone promising otherwise is selling a timeline they cannot control. Paid acquisition answers the immediate question honestly.

If the category genuinely has no search demand, and some do not, then building for search is building for an audience that is not looking. That is worth establishing with data before a budget is committed rather than after.

Who does the off-site half, what you keep, and where a person signs

The off-site half of this is the part most businesses can genuinely begin themselves, and it costs attention rather than money. Getting listed accurately where your category is already discussed, being named somewhere that resolves to the same organisation, and correcting the two profiles that describe you differently are all jobs for your own team, and they move generative visibility further than a month of on-page work does. What is harder to hold in-house is the entity consistency audit across a hundred pages, and the measurement, because taking the same reading repeatedly and acting on what moved is exactly the work that stops the week somebody is busy.

Nothing here publishes on its own. Every page change, every correction to a profile and every reply to a review is prepared and then waits for human approval, because a page in your name and a reply carrying a customer's name have consequences no workflow can weigh. Where the research finds something genuinely ambiguous, such as whether a claim about your business is still true, it hands it to a person rather than resolving it in whichever direction reads better.

Ownership matters more here than in most of this work, because the effect compounds and a supplier holding it holds the compounding with it. The pages sit on your domain. The profiles are registered in your name. The keyword research, the entity audit and the rankings history live in accounts under your logins. On the build and train route you own the system outright at the end of the first phase, and on the rented route you are renting output by agreement rather than by accident, which is why the two are set out separately above instead of blurred together.

Ali owns marketing and sales at Wobble and Haad owns growth and client solutions. Moiz Khan owns automation architecture and decides what any of these workers may do unsupervised. The company works from Karachi, bills month to month and is answerable for what it operates, across 25 engagements in six countries. The reach quoted higher up this page, 3.97 million lifetime impressions for Palm Berries and more than 987,000 for Center for Sight, comes from accounts Wobble operates and sits on the work page with the rest of the numbers.

Common questions

What is the difference between SEO, AEO and GEO?

SEO is ranking in a list of links. AEO, answer engine optimisation, is being the passage extracted into a direct answer, which rewards answering one question completely in one place. GEO, generative engine optimisation, is being named when an assistant recommends companies in your category, and that is largely decided by what the web says about you rather than by what your own site says.

How do you get mentioned by ChatGPT or Gemini?

By being present in the sources those systems read. That means external mentions, listings, discussions and coverage that associate your name with the category, plus an entity description consistent enough that separate mentions are recognised as the same company. A technically excellent website with no external footprint does not appear in these answers.

Does schema markup help with AI search visibility?

It helps machines parse a page accurately, which is worth having. It is not a visibility lever by itself, and several structured data types that were once useful no longer produce any search feature at all. Emitting a type is not the same as earning something, and any provider counting schema types as a result is measuring the wrong thing.

How long does AI search visibility take?

Technical and entity fixes register in weeks because they change how existing pages are read. Content depth and external presence take months, because they depend on accumulation rather than on a change you can deploy. Anyone quoting a fixed date is quoting a sales cycle, not a search cycle.

Is SEO dead now that AI answers questions?

No, and the reason is mechanical. Answer engines and assistants read the pages that rank. Losing search visibility removes you from the source material the answers are built from, so the two are connected rather than in competition.

How do you measure whether AI search work is doing anything?

Rankings and impressions still matter as inputs. On top of them: whether the business is named in generated answers, on which queries, alongside which competitors, and how that changes over time. Without that second half, a search programme in the current environment is reporting on half the surface it is meant to be winning.

See where this applies to your business

The AI Readiness Call is a short, free conversation about where automation would actually pay back in your business. The call is free. The diagnosis is not.

Book AI Readiness Call