Your AI visibility retainer is up for renewal. What should you check before you sign another term?

What is a renewal review, and when should it start?

A renewal review is a short, structured check of whether a year of AI visibility work moved the number you were sold, measured the way it was measured at the start. It ends in one of four decisions: renew as is, renegotiate, re-tender or bring the work in-house. It should start before the contract's notice date, because that date decides your options.

Many retainers renew by themselves unless one side gives notice, often 30 to 90 days before the term ends. Read the termination clause first and put the notice date in a calendar. If the review cannot finish before it, ask the agency in writing for a short extension of the notice window. A provider confident in its results will usually agree.

This is also the first renewal many buyers face in this category. In AgencyAnalytics' 2026 benchmarks survey of 494 agency professionals, 64% of them in the US, fielded February–April 2026, 66% reported increased client demand for AI search services.¹ AgencyAnalytics sells reporting software to agencies. Many of the contracts signed on that demand are now reaching their first anniversary.

Prior work and a disclosure

Sigzen AI sells AI visibility programmes at published prices, with a six-month minimum and part of the fee at risk on the upper tiers, so we have an interest in how renewals are judged.³ The checks below work whoever you buy from. Our earlier guide to what a retainer should deliver each month covers the signing stage; this post covers the moment a year later when you decide whether to sign again.

Others have looked at how buyers judge this work. Digiday reported in May 2026 that marketers were questioning the price of AI visibility tools as results varied between runs and between tools.⁴ That scepticism belongs in the renewal meeting, and it is best answered with a method you can inspect.

Gather the evidence before the meeting

A renewal decision is only as good as the file you bring to it, and most of that file should already exist. Ask the agency for it in writing at least four weeks before the notice date, so that missing items show up while there is still time to chase them.

Ask for six things. The prompt set as it stood at the start and as it stands now, with every change dated. The baseline report and every monthly report since. The raw answers behind each report, with the engine, country and capture date. The list of published assets and the URLs they live on. The invoices. Any change orders or scope notes agreed along the way.

Two of those items matter more than the rest. The raw answers let you check any number in any report, and the change log tells you whether the reports can be compared at all. If the agency's tool is the only place the answers exist, ask for an export now, while the relationship is friendly.

A renewal review on a 60-day notice period

1. Day 1

   Read the termination clause, note the notice date, request the evidence file in writing.

2. Week 2

   Evidence arrives; check the prompt set and change log for breaks in comparability.

3. Week 3

   Re-run the baseline protocol yourself or through a third party on a sample of prompts.

4. Week 4

   Reconcile deliverables against invoices; review ownership of data and assets.

5. Week 5

   Decide: renew, renegotiate, re-tender or bring in-house. Tell the agency before the notice date.

Is the number still measured the way it was at the start?

Only if nothing in the measurement changed, or every change was logged and re-baselined. A share of answer from January and one from December describe the same thing only when the prompts, engines, country, language, session type and run schedule match. When any of those moved, the trend line in the report may reflect the method rather than the work.

Prompt sets drift for honest reasons. Products launch, a market is added, a question goes stale. The problem comes when prompts are swapped without a note, because new prompts can be easier to win than the ones they replaced. Compare the first and last versions of the set side by side and count how many prompts survived the year unchanged.

Repeat runs need the same care. MaxAEO, which sells AI visibility measurement, published guidance treating repeat runs of one prompt as clustered: they resemble each other, so extra repeats add less information than new prompts would.⁸ If the run schedule changed during the year, a rise in the reported number can come from counting differently rather than from being named more often.

Separate the agency's effect from the engines' changes

Engines changed during every retainer that ran through 2026. OpenAI moved paid ChatGPT plans to GPT-6 Sol from 7 October 2026 and Free and Go users to GPT-6 Luna from 8 October.⁶ A change like that can lift or cut share of answer for every brand in a category in the same week, with no work done by anyone.

How often an engine searches the web matters too. Ahrefs, which sells SEO and AI visibility tools, studied 66.6 million free, logged-out ChatGPT responses from August and September 2026, across all countries, and 14.4% triggered a web search.⁷ Work on your site and third-party pages reaches the searched answers first. A report that splits results by engine and marks model changes on the chart lets you see which movements the agency can claim.

Our view: ask for the same chart for two or three named competitors, measured on the same prompts. If everyone in the category rose by a similar amount in the same month, the engine moved, not the retainer. Our guide to measuring before and after a model release sets out that comparison step by step.

Did the result move on the prompts your buyers ask?

That is the question the renewal turns on, and it needs the prompt set split in two. Unbranded buyer questions, the ones that do not contain your name, are where shortlists form and where rivals compete. Branded prompts tell you whether the engines describe you correctly, which matters, but a gain there says little about new buyers finding you.

Branded results rise with little help. In Ahrefs' US desktop data, 73.3% of branded keywords triggered a Google AI Overview in September 2026, as a monthly average.⁹ Profound, which sells AI visibility tools, analysed 11.84 billion citations across eight engines from 16 April to 16 July 2026 and found that about 57% went to company-owned sites.¹⁰ When a prompt names you, your own site is the obvious source, so a branded gain is cheap to report.

Reading a year of results before you renew

Evidence you can check →

Clean method, flat resultRenegotiate scope or re-tender; the agency can at least show you why.

Clean method, unbranded gainRenew, and fix the terms that were loose.

Thin evidence, flat resultRe-tender, and ask for your data before you give notice.

Thin evidence, reported gainRe-run the baseline yourself before you believe the gain.

Movement on unbranded buyer prompts →

Re-run part of the baseline yourself

You do not need the agency's tool to spot-check its numbers. Take a sample of the unbranded prompts, run them in clean signed-out sessions on the named engines, in the agreed country and language, on two or three separate days, and note which brands are named. Compare your tally with the agency's latest month for the same prompts.

A sample will not match the report exactly, because answers vary between runs. It should land in the same range. A large gap in the agency's favour is the point to ask how the count was made. Our one-hour self-check sets out a manual method, and our measurement spec lists what any vendor's count should document.

Reconcile what you received against what you paid for

The contract listed deliverables, and the invoices charged for them, so the third check lines the two up month by month. This is the least contested part of the review, because it rests on documents both sides already hold. It also tends to surface the most specific points for renegotiation.

Build a simple table: each month, the deliverables promised, the deliverables shipped with links, and the invoice. Look for three patterns. Months where assets were promised and not published. Work that was published but never appeared in the reports. And scope that grew informally, which you may want written into the next contract or removed from it.

Illustrative. Reconciliation checklist from the Sigzen AI editorial team; adjust the rows to your contract's deliverables.

Our view: a missed deliverable is a negotiating point, not automatically a reason to leave. Agencies under-deliver for reasons that are sometimes yours, such as slow sign-off or missing access. What matters is whether the reports said so at the time or left it out.

Which parts of the work do you keep if you leave?

The parts the contract assigns to you, and the parts you can export before notice is given. Read the intellectual property and data clauses with four items in mind: the prompt set, the raw answers, the published assets and the accounts. If the contract is silent on any of them, ask now, while both sides still expect to keep working together.

The prompt set is the hardest to rebuild. It encodes a year of decisions about which buyer questions matter, in which countries and languages. If it lives only in the agency's tool, a new provider starts from scratch and your trend line breaks. Ask for the full set, with every version dated, as a plain file you hold.

Raw answers come next. Without them, no new provider can show that its first month compares with the old one, and you lose the ability to check past reports. Assets usually live on your own site, but off-site profiles, directory listings and content accounts are sometimes registered to the agency. Move them into your name before notice, not after.

If the agency reports through a third-party tool, ask what happens to your history if that tool changes owner or terms. Our note on tool acquisitions and contract terms lists the clauses to ask for.

What are the four ways a renewal can end?

Renew, renegotiate, re-tender or bring the work in-house. The evidence from the first three checks usually points to one of them. A clean method and a gain on unbranded prompts favours renewal. A clean method with a flat result favours renegotiation or re-tender. Thin evidence of any result favours a re-tender.

Renew or renegotiate

Renewing as is suits a contract that worked and was well written. Renegotiating suits one that worked but had loose terms: a prompt set nobody froze, no raw answers, no fee at risk. Use the review's findings as the agenda. Each gap you found becomes a clause in the next term, and a provider that delivered will rarely refuse to write down what it already did.

Re-tender

A re-tender gives you a market price for the next term, and it is easier the second time because you now hold a baseline. The market is crowded. G2 listed more than 150 products in its answer engine optimisation category by January 2026, up from 7 in March 2025.² Invite the incumbent to bid as well, on the same brief. Our RFP question list covers what to ask the bidders.

Bring it in-house

Bringing the work in-house means splitting it into measurement and the work that moves the number. Measurement can run on a tool subscription. Content, site fixes and off-site work need people with time. Across the 13 listed agencies pricing in US dollars that we recorded on 5 and 6 October 2026, the median monthly entry price was $3,400, and listed tools had a median entry tier of $79.⁵ The gap between those two prices is mostly the cost of the people.

The market you renew into

More demand, more vendors, published prices

66%of agency professionals reported increased client demand for AI search servicesAgencyAnalytics, Feb–Apr 2026 ¹

150+products in G2's answer engine optimisation category, up from 7G2, Mar 2025–Jan 2026 ²

$3,400median monthly entry price of listed agencies pricing in US dollarsSigzen AI, 5–6 Oct 2026 ⁵

Our view: in-house measurement with outside help for specific projects works for firms with a content team and one or two markets. For several countries and languages, or for off-site work that depends on press and trade relationships, the people cost usually decides it the other way.

Clauses to change in the next term

Most renewal reviews end with a list of terms that were loose the first time. Write them into the next contract before you sign, whichever provider you choose. Each one closes a gap that made the first year harder to judge.

Six clauses for the next term

1. Frozen prompt set

   Fixed for each quarter, with every change dated and the changed prompts re-baselined.

2. Written protocol

   Named engines, clean signed-out sessions, fixed country and language, runs on separate days.

3. Raw answers monthly

   Answer text, engine, country and capture time for every prompt, held on your side.

4. Unbranded target

   The headline target is share of answer on unbranded buyer prompts, by engine.

5. Ownership and export

   Prompt set, raw answers, assets and accounts are yours, exportable within days of a request.

6. Notice and review dates

   A review meeting booked well before the notice date, written into the contract.

Consider tying part of the fee to the result as well. A fixed fee with a defined share at risk against a measured share of answer pays the agency for the work and rewards it only for movement you can check. Our note on pay-per-citation pricing compares that structure with a price per link.

Our programme is one version of it. On Growth and Scale, 20% of the monthly fee is at risk against the share-of-answer target set at the audit, and is invoiced only for a quarter in which the target is met, as on our pricing page.³ Every tier refreshes the prompt set quarterly. Other designs exist; what matters is that the target and the method are written down before the term starts.

A renewal review you can run in five weeks

Start with the notice date and work backwards. Ask for the evidence file in writing, check the prompt set and change log for breaks, re-run a sample of unbranded prompts yourself, reconcile deliverables against invoices, and confirm what you own. Then decide among the four outcomes, with the findings as your agenda.

The review is also the first piece of work for the next term. Its baseline, its frozen prompt set and its list of loose clauses become the starting point, whoever does the work. A buyer who renews with those in hand is in a stronger position than one who signed the first contract on a pitch.

Our view: the best renewal outcome is a contract you could hand to a different provider without losing your history. If the prompt set, the raw answers and the method are yours and written down, renewing becomes a choice about quality and price, and the cost of switching stops deciding it for you.