Most small businesses do not need more SEO activity. They need SEO they can check. If an agency says traffic is up, you should be able to see where it came from. If they say content is working, you should be able to tie it to enquiries, calls, demos or sales. If they say your business is appearing in AI answers, that should be measurable too.
That is the standard we use at Seonis. We built our platform for companies with websites in their own language, not in English, and for agencies serving those companies. Good SEO for agencies is not about impressive decks or vague monthly reports. It is about a repeatable process, clear controls, and outputs a client can verify in Search Console, in their CMS, in their lead data and in AI answer tracking.
What good SEO for agencies looks like for small business clients
For an SME client, good agency SEO starts with the basics. The agency should know what the business sells, where it sells it, what language buyers use, and what counts as a lead. Without that, everything else is noise.
A proper starting point usually includes:
- access to Google Search Console
- access to the website CMS, whether that is WordPress, Shopify, Webflow or Ghost
- a list of services or product categories that actually matter commercially
- the target countries and languages
- the client’s current source of leads, such as forms, phone calls, booked meetings or checkout sales
The first deliverable should not be fifty pages of generic advice. It should be a prioritised plan. That means identifying which pages need fixing, which topics deserve new content, and which technical issues are blocking indexation or relevance.
For a small business client, we think an agency should be able to show delivery in five areas.
First, technical basics. Pages should be indexable. Canonicals should make sense. Redirects should not be broken. Important pages should load properly on mobile. International sites should use the right language targeting approach. If a company has separate country or language sections, the structure needs to be consistent. In the EU this often means one language serving several markets, while in the UK and Ireland the issue is usually less about multiple languages and more about local service intent.
Second, keyword research tied to buyer intent. Not every search term matters. A local accounting firm in Poland does not need traffic from broad informational terms if the real business comes from searches about a named service in a named city. A German B2B manufacturer may need product specification content, not blog posts about general industry trends. Agencies should be able to explain why each target term matters and what page is meant to rank for it. We cover that process in our guide to finding keywords that bring the right buyers.
Third, native-language content. This is where many agency programmes fail. They publish a steady stream of text, but it is translated from English briefs, written in flat language, and disconnected from what local buyers actually type. The output looks busy and reports well, but it does not convert.
Fourth, authority building. For some sectors, strong on-page work is not enough. The agency needs a sensible way to improve authority through links and mentions. Not random directory spam. Not bulk placements on irrelevant sites. The method should be documented, controllable and easy to audit. If an agency is outsourcing this, the client should know the rules.
Fifth, reporting that a non-specialist can verify. A client should not need to trust an agency blindly. They should be able to open Search Console and see which queries and pages gained impressions and clicks. They should be able to inspect published articles in the CMS. They should be able to compare lead volume and quality before and after.
That is what good SEO for agencies looks like when the client is a real business with limited time and budget. Not endless activity. Useful work, in the right language, tied to business outcomes.
How agencies should handle content for non-English websites
For non-English websites, translated English content is usually the wrong starting point.
The problem is not only style. It is keyword mismatch. People do not search in neat one-to-one translations. They use local phrasing, local product names, local abbreviations and local intent patterns. A phrase that looks correct to a translator may have little or no search demand. Worse, it may signal the wrong intent.
Take service pages. In one market, buyers may search for the service name plus city. In another, they may search for a problem, not the service category. In another, they may use an English loanword. There is no shortcut here. Agencies need keyword research in the site’s own language.
That means the process should look like this:
- research terms natively, not by translating an English list
- group keywords by intent and page type
- check what already ranks in that market
- write in the language buyers actually use
- review the draft through a native-language quality gate before publishing
At Seonis, that is exactly how we work. We research in the website’s language, generate content natively rather than translating from English, and check each piece through a language gate before it goes live. That matters because low-quality multilingual content is easy to spot, and users spot it before Google does.
A lot of agencies have learned this the hard way. They scaled content production using English-centred workflows, then found that pages indexed but did not attract the right clicks. Or they got traffic that never converted because the article answered a different question from the one local buyers had in mind.
A better content brief for a non-English site includes:
- the target query set in the local language
- the search intent
- the page type, such as article, category page, service page or location page
- internal links to relevant existing pages
- any terms, claims or regulated wording the client needs to avoid
- the conversion goal
Regulated sectors need extra care. In the EU generally, marketing rules differ by country and by industry, especially in finance, health and legal services. In Germany, for example, wording around guarantees and comparative claims can create problems faster than in less tightly interpreted markets. In the UK, the legal framework is separate from the EU, and sector-specific regulators may matter more than EU-wide practice. An agency writing at scale needs controls for this, not just a content calendar.
For agencies handling multilingual accounts, the safest model is a standard process with local-language execution. We wrote more about that in our piece on SEO tools agencies can trust for multilingual client work.
What an agency needs to measure beyond rankings
Rankings matter, but they are not enough. A client can rank fifth for a term that brings no business. They can also gain business from terms that never show up in a simplified rank tracker because of location, device, personalisation or AI answer behaviour.
The first place to look is Search Console. An agency should be able to show, at minimum:
- which queries gained impressions
- which pages gained clicks
- the click-through rate trend on key pages
- average position changes for meaningful query groups
- indexation status for newly published content
For small business clients, page-level reporting is often more useful than a giant keyword list. If a new service article starts getting impressions for the right terms, then clicks, then enquiries, that is progress the client can understand.
But Search Console has limits. It does not tell you whether the leads were any good. Agencies need to connect search performance to outcomes. For lead generation businesses, that means tracking form submissions, calls, booked meetings or qualified enquiries. For e-commerce, it means revenue and assisted conversions.
This sounds obvious, yet many SEO reports still stop at visibility metrics. We think that is where agencies lose trust. If a client asks, “Which pieces brought actual enquiries?”, there should be an answer. Our article on how to judge SEO articles that bring real enquiries sets out the standard we use.
There is now a third layer as well, AI answer visibility.
Buyers increasingly ask ChatGPT, Gemini, Perplexity, Claude and Grok for suppliers, comparisons and recommendations. If your business is never named in those answers, that matters. If it starts appearing, that matters too. Agencies need to treat this as something measurable, not a talking point.
At Seonis, we track whether these systems mention the business when buyers ask relevant commercial questions. This is not the same as traditional ranking data, and it is not perfect. AI systems can vary by prompt, location, account state and time. But that is not a reason to ignore them. It is a reason to monitor them consistently and report the trend.
The numbers we care about are simple:
- are we being mentioned for the right commercial prompts
- how often are we being mentioned over time
- which topics trigger mentions
- which competitors are appearing instead
This gives agencies a way to discuss AI visibility without pretending it works exactly like search rankings. It does not.
How agencies can scale SEO without turning it into generic AI output
Scaling SEO across many client accounts is not the problem. Scaling it without losing relevance is the problem.
Most poor agency output comes from skipping one of five stages: research, drafting, checking, publishing or review. If the workflow is weak at any stage, volume makes the weakness worse.
A scalable process should be boring in the best sense. Repeatable. Auditable. Hard to fake.
We use a nightly autonomous workflow, but the principles are simple enough for any agency model.
Research comes first. Each account needs its own keyword set, in the client’s language, mapped to real pages and commercial priorities. No shared master list copied across sectors.
Drafting comes next. The draft should be based on that research, written for the intended page type, and structured to support internal linking and conversion. The aim is not to sound clever. The aim is to answer the search intent better than what is already ranking.
Then checking. This is where most AI-heavy content systems fail. They can draft quickly, but they cannot be trusted to know whether a phrase sounds natural in Latvian, Polish or German, or whether it accidentally uses the wrong term for a regulated service. We put every piece through a language gate before publishing. Agencies should have an equivalent control, whether in-house or through a platform.
Publishing should also be part of the system, not a manual bottleneck. If the agency needs three logins, two spreadsheets and a developer ticket to get one article live, scale will collapse. The workflow should support direct publishing into WordPress, Shopify, Webflow or Ghost with the metadata, links and formatting intact.
Then review. Not every page will work. Some will gain impressions but not clicks. Some will attract the wrong audience. Some will need stronger internal links or a sharper title. Review is where scaled SEO becomes accountable.
Authority work also needs a process. Agencies often struggle here because link building is labour-heavy and easy to make messy. We use a members’ exchange model priced in credits, so the rules are visible and the work can be managed consistently across accounts. For agencies that want a structured model rather than ad hoc outreach, our explanation of how white label link building works for smaller agencies sets out the practical detail.
The point is not that every agency must use our exact setup. The point is that SEO for agencies only scales when each stage has controls. If you skip the local-language research, you get irrelevant topics. If you skip the language gate, you get generic AI text. If you skip publishing integration, content sits in drafts. If you skip review, bad pages stay bad.
What to ask before you hire an agency for SEO
If you are hiring an agency, ask questions that force specifics.
Start with the content process.
Ask:
- Do you research keywords in our language, or translate them from English?
- Who writes the content?
- Who checks it before it is published?
- Can we see examples on sites like ours?
If the answers stay vague, that is a warning sign.
Then ask about measurement.
- What will you show us in Search Console each month?
- How will you connect SEO work to leads or sales?
- Do you report at page level?
- How do you measure visibility in ChatGPT, Gemini, Perplexity, Claude and Grok?
If the agency talks only about rankings, they are leaving out too much.
Ask about publishing and ownership.
- Will content be published directly to our site?
- Which CMS platforms do you support, such as WordPress, Shopify, Webflow or Ghost?
- Do we own the content and the reporting data?
- What happens if we stop working together?
These are basic questions, but many clients only discover the answers after months of spend.
Ask about links and authority.
- How do you build links?
- Can you explain the quality controls?
- Will we be able to review placements?
- Are there any methods you avoid completely?
A good agency should be able to answer this plainly. If the method sounds secretive, it is usually because the quality is weak.
Ask about operational discipline.
- What is the monthly output?
- How are topics approved?
- What is the review cycle?
- How do you handle regulated sectors or restricted claims?
- Who is accountable if content is wrong?
And ask one question many buyers forget:
- What should we be able to verify ourselves after 90 days?
That question changes the conversation. It moves the agency away from promises and towards evidence.
For agencies looking to offer this standard to their own clients, our partner programme for agencies is built around exactly that model. Native-language research and writing, publishing, authority building, and reporting the client can check in their own language.
The businesses that win from SEO are not always the ones spending the most. They are often the ones with the clearest process and the fewest blind spots. If you run or hire SEO for agencies, that is the benchmark to use. Can the client verify what was done, where it was published, what changed in Search Console, what leads came in, and whether the business is now being named in AI answers?
If the answer is no, the reporting may look tidy, but the work is not under control.