Enterprise Fully Managed SEO, in the Age of AI
Fully managed SEO is a buying model rather than a technology: an outside team owns the programme and the buyer pays a recurring fee for the outcome. The real question is never which agency. It is build, buy, or hire — and Google's first page answers none of it, because every page ranking there is sold by one of the three options.
By Vijay Vasu, Founder, Indexable. Published September 9, 2026.
How we measured. A live Ahrefs SERP fetch for “fully managed seo”, US, top 10 organic positions, 9 September 2026. Hiring economics from a corpus of 29 public senior search postings collected June to August 2026, of which 16 disclosed an annual band; that sample is self-selected and mixes geographies and seniorities, so treat it as a reference point rather than a market benchmark. Search Console figures come from the indexableai.com property (Google Search Console, 2026), 90 days to 8 September 2026, 1,063 matching query rows. Per-engine share of voice from Ahrefs Brand Radar on a panel of 106 custom prompts, US, 7 September 2026, excluding Claude, whose response count was too small to produce a real distribution. One domain, one results page, one panel — a case study, not a law.
- The AI Overview for “fully managed seo” is assembled from seven agency service pages, and every organic result on that page is an agency selling a retainer or a directory listing agencies (Ahrefs, 2026).
- Google's People Also Ask block asks “How much should I pay for SEO services?” and “Can I manage my own SEO?” (Ahrefs, 2026) — the two questions nobody on that page is disinterested enough to answer.
- The median advertised band across 29 public senior search postings collected June to August 2026 was $176K–$221K, from the 16 that disclosed one (Indexable, 2026).
- 17 of those 29 postings demanded eight or more distinct specialisms, and 10 of the 17 named no team (Indexable, 2026).
- webfx now titles its services page “SEO Services: Get Discovered in AI & Traditional Search” (Ahrefs, 2026). The incumbents are repositioning inside their own title tags.
- Scope written around rankings misses most of the outcome: 18 of our pages sat in Google's top 10 and 3 were cited by an AI engine (Indexable, 2026).
- Share of voice on our own prompt panel ran from 58.57% on Copilot to 1.17% on Gemini in one day, a 57.4-point spread (Indexable, 2026).
- Citations lag publication by a median of 6.81 days and a 90th percentile of 37.10 days (Profound, 2026).
What is fully managed SEO in the age of AI?
Fully managed SEO is an arrangement in which an external team owns the whole programme — strategy, technical remediation, content and reporting — and the client pays a recurring fee for it. The model has not changed. The scope of work it has to cover has.
The traditional scope was written around one outcome: rank, and therefore get the click. That outcome is still real and still the majority of measurable traffic on most estates. It now sits alongside a second outcome the same scope rarely mentions — whether a retrieval system can lift a passage out of the page and attribute it without a visit ever occurring.
The two need different work. Ranking work is about competitiveness: architecture, authority, depth. Retrieval work is about extractability: self-contained claims, question-shaped headings, visible dates, structured data treated as a contract rather than a decoration.
The commercial consequence is specific to this topic. A managed engagement is defined by its statement of work, and most statements of work in this market were drafted when there was one outcome to buy. You should read yours against both columns before renewing it.
Why does the results page never price the alternative?
Because nobody on it is disinterested. The AI Overview for “fully managed seo” is assembled from seven agency service pages, and the organic results underneath are agencies plus a directory whose business is listing agencies (Ahrefs, 2026).
The composition is worth naming. Position 2 is seo.co/services/ at domain rating 74. Position 4 is wordlift.io/seo-management-service/, also 74. Position 9 is clutch.co/us/seo-firms, a directory, at domain rating 91. Position 10 is webfx.com/seo/services/ at domain rating 89 with 339 referring domains (Ahrefs, 2026).
Two results on that page report traffic figures we are deliberately not publishing: tens of thousands of monthly visits against one to three referring domains, which is the signature of a tool attributing site-wide traffic to a single URL, or of a local doorway pattern. We flag them as anomalous and set them aside.
Now read Google's own People Also Ask block. It asks “How much should I pay for SEO services?” and “Can I manage my own SEO?” (Ahrefs, 2026). Google is putting the build-versus-buy question to the buyer, and the page answers with ten vendors. That gap is the opportunity in this topic.
The incumbents are already repositioning around AI search
The clearest signal on this results page is not in anybody's marketing copy. It is in a title tag. webfx now titles its SEO services page “SEO Services: Get Discovered in AI & Traditional Search” (Ahrefs, 2026), and that page sits at position 10 with domain rating 89 and 339 referring domains — the most heavily linked page in the set.
A title tag is not a blog post. It is the most contested sentence on a commercial page, rewritten rarely, and rewritten only once the positioning behind it has moved. When a well-linked supplier puts “Get Discovered in AI” ahead of the traditional promise, the pitch has changed before the results page has.
What a title tag cannot tell you is whether delivery changed with it. That is the diligence question, answerable in one meeting: ask what data source the AI numbers come from, ask for a per-engine breakdown with a stated sample size and date, and ask whether server logs are in scope.
One honest limit: this is one title tag, on one page, on one day, so we are reading a signal rather than a strategy.
See which engines can actually retrieve you
The free AI search audit returns the per-engine split described above — with the sample size and date attached, so you can audit it.
What does the hire option actually cost?
More than most budget models assume, and scope is the reason rather than seniority. Across 29 public senior search postings collected June to August 2026, the median advertised band was $176K–$221K, from the 16 postings that disclosed one (Indexable, 2026).
The scope numbers are the more useful half. 17 of the 29 postings demanded eight or more distinct specialisms in a single hire, and 10 of those 17 named no team at all (Indexable, 2026). 11 of the 29 wanted technical, content, analytics and engineering competence in one person (Indexable, 2026).
Three limits. The corpus is public postings and self-selected, it mixes geographies and seniority, and an advertised band is an advertisement, not a paid salary.
| Option | What the buyer is buying | The number we can put on it | Where it fails |
|---|---|---|---|
| Hire | One person's full-time judgement | Median advertised band $176K–$221K, 29 public postings, June to August 2026 (Indexable, 2026) | 17 of 29 want 8+ specialisms; 10 of those name no team (Indexable, 2026) |
| Buy a managed engagement | Capacity, and someone else's process | No public figure on this results page — the price question is asked in People Also Ask and answered with vendors (Ahrefs, 2026) | Scope usually written around rankings alone |
| Build in-house | Ownership of the measurement stack | No single figure; specialism spread 4 to 12 per posting, median 8 (Indexable, 2026) | Slowest to stand up; per-engine and log measurement is the scarce skill |
What does “managed” have to cover now?
Three things that were not in a 2022 statement of work, and none appear in a standard rankings-and-traffic report. Each is a contractual question rather than a technical one, so you should write all three into the scope before the first invoice.
Server-log segmentation. Client-side analytics record zero AI agent fetches, because those agents do not execute the JavaScript that fires the analytics beacon (Osmani, 2026). An engagement reporting from a client-side property alone cannot see the retrieval layer. Ask who parses the logs.
Per-engine visibility. On our own prompt panel, share of voice ran from 58.57% on Copilot to 1.17% on Gemini on the same day — a 57.4-point spread across one brand, one panel, one day (Indexable, 2026). A blended score averages that away and hides which engine you lost. We excluded Claude deliberately: its response count was too small to produce a real distribution, and publishing it would have manufactured a zero.
A 45-day evaluation window. Citations lag publication by a median of 6.81 days and a 90th percentile of 37.10 days in Profound's 2026 analysis of agent logs across roughly 900 pages (Profound, 2026) — Profound's measurement, not ours. A monthly cycle grades much of the work before it can land.
What decision does the executive now own?
Three, and none of them is a supplier decision.
Which outcome the contract is written around. On our own estate, 18 pages sat in Google's top 10 and 3 were cited by an AI engine — 15 ranked and were never retrieved (Indexable, 2026). The inverse held too: our two most-cited pages rank at positions 30.5 and 28.1, which is page three (Indexable, 2026). A scope that pays for the first column and reports on it exclusively looks successful while the second decays.
Who owns the measurement stack at termination. If the log pipeline, the prompt panel and the citation history live in a supplier's account, the buyer is renting the evidence and restarts from zero at renewal. Decide at signature whether instrumentation is a deliverable you keep or a dependency you rent.
How long you wait before judging it. Set the review at 45 days rather than 30, given a 90th-percentile citation lag of 37.10 days (Profound, 2026). Then get written agreement, in advance, that a click decline alongside a citation rise is an acceptable result.
What are the anti-patterns?
Five recur in this market, and each has a correction you can apply during procurement.
Buying on the deck rather than the measurement stack. Ask what data source the AI numbers come from. If the answer is a client-side analytics property, there is no data behind them, because those properties record zero agent fetches (Osmani, 2026).
Comparing suppliers only to each other. The real comparison set is build, buy and hire. Three retainers weighed against three retainers is a choice inside one column.
Scoping the engagement on head terms. On our domain, head terms sit between positions 76.8 and 89.0 while question-shaped queries sit between 1.9 and 5.3 (Indexable, 2026). A deliverable list built on head terms points the budget at the part of demand our estate performs worst on.
Reading a single blended AI visibility score. One number across engines hid 57.4 points of spread on our panel in a single day (Indexable, 2026). Insist on the per-engine split with a stated sample size and date.
Judging the programme on a monthly cycle. With a 90th-percentile citation lag of 37.10 days (Profound, 2026), a 30-day verdict is a coin toss presented as a decision.
How do you run the build-versus-buy-versus-hire comparison?
Seven steps, and the first two can be done from a spreadsheet this week.
- Step 1 — price all three columns before you shortlist. Start by putting the hire, the managed engagement and the in-house build side by side on one page, with a number in each. Use that page as the front sheet of the brief.
- Step 2 — count the specialisms your problem needs. The median posting in our corpus asked for 8, across a range of 4 to 12 (Indexable, 2026). Then, check whether your preferred option supplies all of them.
- Step 3 — ask every supplier where the AI numbers come from. You should require a named data source in writing. A client-side analytics property cannot see agent fetches (Osmani, 2026), so that answer disqualifies the number.
- Step 4 — require per-engine reporting with a stated sample size and date. Apply the same rule internally, so both sides are auditable.
- Step 5 — implement server-log segmentation before the engagement starts. You can do this without any supplier, and it gives you a pre-period nobody else controls.
- Step 6 — schedule the first review at 45 days, not 30, given a 90th-percentile citation lag of 37.10 days (Profound, 2026).
- Step 7 — write instrumentation ownership into the contract. Next, name in the same clause who holds the prompt panel and the citation history at termination.
If you cannot complete step 1 for all three columns, you are not choosing between options. You are choosing between vendors.
In summary
The market for fully managed SEO sells one column of a three-column decision, and its results page is composed entirely of parties who profit from that column (Ahrefs, 2026). The buyer's choice is build, buy or hire, and only one of the three carries a public number: a median advertised band of $176K–$221K across 29 senior postings (Indexable, 2026).
Whichever column you pick, the scope has moved. Server-log segmentation, per-engine visibility with a stated sample size and date, and a 45-day evaluation window separate a programme you can audit from a report you take on trust.
The next step is step 1 above, and it costs an afternoon: put the hire, the managed engagement and the in-house build on one page with a number against each. If two of the three cells come back empty, that is the finding.
The Managed SEO Procurement Check
- Have you written a number against all three columns — hire, buy, build — before shortlisting anybody?
- How many distinct specialisms does your problem actually need?
- (Pause point — if item 1 has two empty cells, stop. You are comparing vendors, not options.)
- Can every supplier name the data source behind their AI numbers, in writing?
- Does the reporting break down per engine, with a stated sample size and date?
- Are server logs in scope, and is it written down who parses them?
- Is the first substantive review scheduled at 45 days, not 30 — with written agreement that a click decline alongside a citation rise is acceptable?
- At termination, who holds the log pipeline, the prompt panel and the citation history?
Scoring — count the empty cells, not the yeses:
- Three columns priced, items 4–8 all yes — Instrumented. You are choosing between options and you will be able to audit whichever you pick. Proceed.
- Three columns priced, one or two of items 4–8 missing — Comparing. The decision is sound; the measurement is not. Fix the gaps before signature, not at renewal.
- Only the "buy" column priced — Vendor-Shopping. This is where the results page leaves everybody. You are running a supplier bake-off inside one third of the decision.
- Items 5 and 6 both no — Unmeasured. Whatever you buy, you will not be able to tell whether it worked. Start there.
Our own baseline, stated honestly. The first time we ran this on ourselves we scored Unmeasured on items 5 and 6 — we had no per-engine reporting and no log segmentation. That is why both are on the list. One domain, one window, one panel: a case study, not a law.
Frequently asked questions
How much should I pay for fully managed SEO?
Price it against the alternatives rather than against other suppliers, because the results page for this term carries no disinterested source (Ahrefs, 2026). The most useful anchor is the hire: across 29 public senior search postings collected June to August 2026, the median advertised band was $176K–$221K from the 16 that disclosed one (Indexable, 2026). That sample is self-selected, so use it to size the decision, not to benchmark a quote.
Can we manage SEO in-house instead?
Sometimes, and the deciding factor is specialism coverage rather than headcount. 17 of 29 public postings we examined demanded eight or more distinct specialisms in one hire, and 10 of those 17 named no team (Indexable, 2026). If your problem needs technical, content, analytics and engineering competence at once — as 11 of the 29 did — one hire is a bottleneck.
How long before a managed engagement shows results?
Longer than a monthly reporting cycle. Citations lag publication by a median of 6.81 days and a 90th percentile of 37.10 days in Profound's 2026 analysis of roughly 900 pages (Profound, 2026) — Profound's measurement, not ours. Set the first substantive review at 45 days, and agree in advance that a click decline alongside a citation rise counts as progress.
Vijay Vasu is the founder of Indexable. SERP figures were pulled live on 9 September 2026; hiring figures come from a 29-posting corpus collected June to August 2026; Search Console figures cover the 90 days to 8 September 2026, and its positions are impression-weighted averages. Verified September 9, 2026.
Related reading
- Enterprise SEO, in the Age of AI — why ranking and retrieval came apart.
- Enterprise SEO pricing — what a budget is accountable for.
- AI visibility agencies compared — how the supplier market describes itself.
Price the third column before you sign
We will run the per-engine visibility split and the ranked-versus-retrieved join against your domain, and send you both, with the sample size and date attached.