Discussion starter: an agency question, posted with an answer from Niraj Raut to open the thread. If you have dealt with this on a site, reply with what you saw, especially where it differs.
- Agency
Is AI search optimisation a new retainer line or the same work repackaged? How are you pricing it without overselling?
Disclosure: Niraj Raut, who posted this answer, runs the SEO consultancy linked at the end of it.
1 reply
-
adminAdmin
Mostly the same work under a new name, with a thin layer of genuinely new work on top. For Google’s AI features, Google’s own guide to generative AI search says that, from Google Search’s perspective, optimising for them is still SEO. The same guide says you don’t need AI text files, special markup or Markdown versions, that there is no requirement to break content into tiny pieces, and that you don’t need to write in a specific way for generative AI search. What is actually new sits around the edges: measuring visibility across several AI platforms, deciding which crawlers and AI features to allow, and for some clients, retail feed work and checking how AI tools describe the brand.
So I would not sell a second retainer. I would price the new layer as a scoped add-on based on its real inputs (tool costs, analyst hours, the number of prompts and platforms tracked), send any content or technical fixes it uncovers into the existing retainer backlog, and write down what the add-on does not promise. A standalone “GEO retainer” that re-bills content and technical work already in scope is a straightforward way to oversell, because the client pays twice for the same hours under a new label.
Which parts are repackaged and which are new
The split decides what you can honestly charge extra for. Google’s guide says its generative AI features are rooted in its core Search ranking and quality systems, so for Google the overlap with SEO is close to complete. Other platforms run their own retrieval. OpenAI, for example, documents OAI-SearchBot for ChatGPT search separately from GPTBot, its training crawler. That creates some new technical and measurement work, but far less than a “new discipline” pitch implies.
Work Already in a sound SEO retainer? What AI search changes Crawling, indexing, rendering, internal links Yes Little for Google. For other platforms, confirm their search crawlers are not blocked in robots.txt or by CDN bot rules. Helpful content, clear answers, first-hand expertise Yes Nothing structural, per Google’s guide. Structured data Yes Google says it isn’t required for generative AI search and there is no special schema.org markup to add. Keep markup that earns rich results; don’t sell “AI schema”. llms.txt, Markdown copies, AI text files No Google says you don’t need them to appear in Google Search. If a client wants one anyway, bill it as a small experiment with no promised result. Visibility tracking across ChatGPT, Gemini, Copilot, Perplexity and Google’s AI features No Genuinely new: prompt design, repeated runs, tooling and interpretation. First-party AI reports in Search Console, Bing and GA4 Partly New reports to add to existing reporting. Low ongoing cost once set up. Crawler and AI feature opt-out decisions Partly New trade-offs: search crawlers versus training crawlers, CDN bot settings, Google’s site-level opt-out. Mentions, reviews, PR, community presence Varies The emphasis shifts, but the work is PR and should be priced as PR. Google says seeking inauthentic mentions across the web isn’t as helpful as it might seem. How AI tools describe the brand Partly Clients now read AI answers about themselves. Checking those answers against the site and third-party profiles is scoped, new work. Retail feeds Sometimes Google has added conversational attributes and AI performance insights to Merchant Center; as of September 2026 the AI performance insights are available for English-language queries in Australia, Canada, India, New Zealand and the US. Feed work for AI shopping can be its own project. The first-party reports deserve a line in any proposal, because they limit what anyone can claim. Search Console’s “Search generative AI performance” reports show impressions for AI Overviews, AI Mode and generative AI features in Discover, with no clicks, CTR or queries. They reached all eligible sites worldwide on 31 August 2026. Bing Webmaster Tools’ AI Performance report shows citations and sampled grounding queries, but no clicks. GA4 added an “AI Assistant” default channel group in May 2026, though traffic without a referrer still lands in Direct. Reading these reports is part of normal reporting, not a reason for a new retainer.
Crawler and opt-out decisions are the other piece of real new work, and they carry business risk. Cloudflare added a “Disallow AI Training” option for crawlers that serve both search and training, while its full block option can remove those crawlers from search as well. Search Console’s site-level “Search generative AI control”, which opts a site out of AI Overviews, AI Mode and AI Overviews in Discover, rolled out to all sites worldwide on 31 August 2026. Advising on those settings takes judgement about the client’s business model, and it is fair to charge for that once, as a review.
Where overselling creeps in
Signs the offer is oversold:
- llms.txt, “AI schema”, chunked rewrites or Markdown mirrors sold as essentials. Google says none is needed for Google Search, and public evidence that they drive citations elsewhere is thin.
- A single “rank” in ChatGPT or Gemini. SparkToro and Gumshoe’s January 2026 research, based on nearly 3,000 runs of 12 prompts across ChatGPT, Claude and Google’s AI, found the tools rarely gave the same list of brands twice. It concluded that ranking-position tracking is unreliable, while visibility percentage across many repeated runs is a reasonable metric.
- Guaranteed citations or recommendations. No platform documents a way to secure them.
- Citations or AI impressions presented as traffic. None of the first-party reports above include clicks.
- Revenue forecasts from AI referrals without the client’s own data. BrightEdge data published in September 2026 still shows AI as a small share of total referrals compared with search, although its methodology is not detailed.
- A page rewrite billed once as SEO and again as GEO.
Pricing models and the risk in each
Model What the client pays for Fits when Overselling risk Fold into the retainer and re-scope No new fee. Some hours move from low-value tasks to AI reporting and brand-accuracy checks. Traffic is mostly Google, and the client mainly wants reassurance and a view of the new reports Low for the client; the risk is unpaid scope creep for you Fixed-fee diagnostic A defined baseline: prompt panel, crawler and CDN review, opt-out advice, brand-accuracy check, ranked gap list The first step for any client who asks Low, if every deliverable is named Monitoring add-on Tool cost plus analyst hours, scaled by prompts, platforms, markets and run frequency Buyers research in AI tools before they contact the client (check this with the sales team) Medium: dashboards can imply more precision than sampling allows Separate “GEO retainer” A new monthly fee covering content, technical and PR work Only when it contains work no other line covers, such as a community or PR program High: double billing and vague deliverables Fees tied to citations A bonus per citation or per point of a visibility score Hard to justify at current measurement quality High: volatile, sampled and loosely linked to revenue A separate line item is not dishonest in itself. If a client holds its AI budget apart from its marketing budget, a clearly scoped monitoring line is easy to fund from it. The problem is a separate line whose deliverables overlap with the retainer, or whose success measure is a score nobody can reproduce.
How I would price the new layer
A builder quoting an energy-efficient renovation can charge for the energy assessment and the certificate, because those are new work. It cannot charge a second time for insulation that was already in the quote. The same logic applies here: price the delta, not the label.
- Start with the client’s question, not the tactic. “When buyers ask AI tools about our category in our markets, are we mentioned, and are we described accurately?” is something a client will pay to answer. “GEO” on its own is not.
- Sell a fixed-fee diagnostic first. Build the prompt panel from real sales and support questions, run each prompt several times per platform, and report visibility as a percentage with the run count stated. Add the crawler and CDN review, the opt-out discussion and a brand-accuracy check.
- Price monitoring on inputs. Prompts, platforms, markets and run frequency drive both tool cost and analyst time. Pass tool costs through at cost or with a stated margin, and charge for interpretation hours. The price then scales with scope rather than with hype.
- Route fixes into the existing backlog. Content, technical and PR tasks that the diagnostic turns up go into the retainer backlog, prioritised against everything else, rather than into a parallel roadmap with its own invoice.
- Write the non-promises into the proposal. No guaranteed citations, no position rankings from single runs, citations reported separately from traffic, and AI referral figures shown with their known gaps.
- Set a review point. After two or three reporting cycles, ask whether the add-on has changed a decision or moved a business metric. If it hasn’t, shrink it to a quarterly check.
Hypothetical example:
A B2B software client on a monthly SEO retainer asks for “GEO”. Instead of a second retainer, the agency proposes a diagnostic of about 20 to 30 hours: 60 prompts drawn from sales-call notes, each run five times on four platforms and reported as visibility percentages; a robots.txt and CDN review covering search and training crawlers; a comparison of how each platform describes pricing and integrations against the site; and a ranked gap list. After that, monitoring costs 4 to 6 analyst hours a month plus the tracking tool at cost. Two items from the gap list (an outdated pricing page and integration names that differ across third-party profiles) go into the existing retainer. The client pays once for the new work and never twice for the same rewrite.
If I were setting this up, I would describe it plainly: SEO for Google’s AI features, plus a measurement and governance module for the other platforms. The module earns its own price because it costs real time and tools. Everything else earns its price inside the existing retainer, and the proposal should say so. When the client’s own Search Console and GA4 figures arrive, an offer priced on the delta will still look fair; one priced on the label will not.
Need help with this on your own site?
Niraj Raut works with businesses in Nepal, Australia, the UK, Europe and the US on technical SEO, ecommerce SEO, local SEO and AI search.
Add a reply Cancel reply
Replies are open to approved members. Log in or apply to join.