AI Search Optimization Services: Two Engines, One Programme

Updated September 2026 · Written and maintained by the Progression Agency strategy team

Search optimization now covers index engines and answer engines. Only two of seven workstreams differ — which is why buying the halves separately costs more and proves less.

On this page · 14 sections

Search optimization now covers two engines
Why the generic search terms stopped meaning Google
What covering both engines actually requires
How a combined programme is sequenced
How the two engines are measured, and why they are never merged
The two opposite mistakes
Where the budget should actually go
What happens to your existing SEO investment
Common objections, answered plainly
Everything else we have written on search, AI and getting found
The two engines, compared on everything that matters
What overlaps, what does not, and what it means for what you buy
How the generic head terms behave now
Two reports, never one score

The short answerSearch optimization now means two engines: the index engines that rank pages among ten links, and the answer engines that compose a written response naming three or four sources. They share indexation, crawl health, architecture, speed, linking and authority. They differ on exactly two things — a second set of crawlers that must be allowed and verified, and a rendering standard stricter than Google’s. One technical workstream, two measurement streams, reported side by side and never merged.

Volumes, CPC and domain authority measured September 2026. AI referral share is cited from the largest published study with a stated denominator.

Search optimization now means two engines, not one — A search optimization service that only covers the first engine is covering most of today’s demand and none of the growth.

Search optimization now covers two engines

A search optimization service used to mean one engine: get a page indexed, get it ranked among ten links, hold the position. That work still carries the overwhelming majority of demand and nothing here replaces it.

What changed is that a second class of engine now sits in front of the first. It does not return a list — it composes a written answer and names three or four sources beneath it. There is no second page and no position eight. A brand is named or it is absent.

The two share an input, which is the whole reason a single programme makes sense: answer engines retrieve from the same web indexes that search optimization builds and maintains. Remove the index work and the answer work has nothing to stand on.

Indexation — Shared. Both engines retrieve from web indexes.

Crawl health — Shared. Broken crawling breaks both.

Architecture — Shared. Discovery matters to both.

Speed — Shared. Helps both, decisive for neither.

Internal linking — Shared. Discovery and context for both.

Topical authority — Shared. Earned slowly, serves both.

What is genuinely shared

Indexation, crawl health, site architecture, page speed, internal linking and real topical authority. That is most of a competent technical programme and it serves both engines without modification.

What diverges

Two things, and only two. A second set of crawlers that must be separately allowed and verified, and a rendering standard considerably stricter than Google’s.

Why that matters commercially

If only two of seven workstreams are answer-engine-specific, buying two suppliers means paying twice for five of them and receiving two reports that disagree about what happened.

What we do about it

One technical workstream, two measurement streams, reported side by side and never merged into a single visibility score.

Why the generic search terms stopped meaning Google

Phrases like engine optimization service and search optimization company do not name an engine. They never did. What changed is what Google returns for them — page one now includes generative and answer engine optimization service pages, in several cases at the very top.

That is not a prediction. On ‘engine optimization service’, a 74,000-a-month term, the second result is a generative engine optimization services page from a domain with an authority score of thirty.

The generic head terms, and what they now return — These are generic, engine-agnostic phrases. Nothing in them says Google, and page one increasingly does not either — GEO and AEO service pages are winning several of them outright.

How the meaning of 'search optimization' changed — This is why we treat the generic head terms as covering both engines: Google itself now answers them that way.

The words did not change

‘Search optimization’ is engine-agnostic and always was. Nothing in the phrase specifies which engine.

What they retrieve did

Commercial intent on these phrases increasingly resolves to service pages covering AI search, because that is what buyers turn out to be asking about.

Why incumbents have not adjusted

Most established agencies still treat AI search as a blog topic rather than a service line, so their service pages do not mention it.

What that leaves open

Generic head terms with very high commercial value, currently won by small specialists. A domain with an authority score of six holds a page-two position on the 74,000-a-month term.

Domain authority winning the generic head terms — A DA-6 domain holds a page-two position on a 74,000-a-month term. The generic head terms are not locked, because most incumbents have not updated what their service pages are about.

What covering both engines actually requires

Twelve things, ten of which sit together comfortably. The two that do not are the errors businesses make when they buy the halves separately: using rank tracking as proof of AI visibility, and paying two suppliers to do the same technical work twice.

Nothing on the list is exotic. The reason most suppliers cover only one column is not difficulty, it is that nobody has asked them to cover the other.

What covering both engines actually requires — Ten of these overlap or sit side by side comfortably. The two reds are what happens when a business buys the two halves separately.

What each kind of supplier actually covers — The two outer bars are the real market. Most buyers end up paying two suppliers for overlapping technical work because neither covers both columns.

AI crawler access — Divergent. Answer engine only. Free to fix.

Scriptless rendering — Divergent. Answer engine is far stricter.

Passage structure — Divergent. Answer engine selects passages, not pages.

Prompt measurement — Divergent. No ranked list exists to track.

Entity agreement — Divergent. Wrong answers are worse than absence.

Corroboration weight — Divergent. Third-party agreement carries more here.

How a combined programme is sequenced

Shared technical foundation first, because it blocks everything else. Then the two cheap answer-engine fixes — crawler access and rendering. Then structure, which helps both. Then index-engine content, where the volume still is. Then corroboration, which serves both at once.

Only two of seven steps are answer-engine-only, and both are front-loaded because they are cheap and binary. That ordering is the entire practical argument for one supplier rather than two.

How a combined programme is sequenced — Only two of these seven steps are answer-engine-only. That is the argument for one supplier rather than two.

Fix crawling — First. Serves both engines, blocks everything if broken.

Allow AI crawlers — First. Free, minutes, binary outcome.

Test rendering — First. The largest single cost variable.

Baseline both — First. Rank baseline and prompt baseline, same week.

Restructure templates — First. Helps extraction and readers together.

Report separately — First. Two streams, never merged.

Week one

Crawl health and AI crawler access. Both baselines recorded in the same week — a rank baseline and a prompt baseline, so neither can be reconstructed favourably later.

Weeks two to four

Rendering. This is where the cost lives, and it is an engineering conversation more often than an agency one.

Weeks three to six

Passage structure across the templates that matter. The rewrite improves extraction and human readability together, which is a rare alignment.

Ongoing

Index-engine content where demand justifies it, and third-party corroboration, which builds authority for one engine and trust for the other simultaneously.

How the two engines are measured, and why they are never merged

The index engine has positions, impressions and clicks. The answer engine has citation frequency, share of voice and sentiment across a fixed prompt set. They go on separate pages of the same report, side by side, never combined into one score.

A blended visibility number is the most common way weak work in one channel hides behind genuine progress in the other. If both engines are on one chart, one of them is decoration.

Rankings — Measured. Index engine. Positions one to ten.

Impressions and clicks — Measured. Index engine. From Search Console.

Citation frequency — Measured. Answer engine. Across a fixed prompt set.

Share of voice — Measured. Answer engine. You versus competitors.

Crawler hits — Measured. Both. From your server logs.

Render coverage — Measured. Answer engine. Templates machines can read.

Index engine measurement

Rank tracking, Search Console impressions and clicks, and the conversion behaviour behind them. Mature, well understood, and not in question here.

Answer engine measurement

A fixed set of buyer questions, re-run with identical wording, with the answers stored verbatim so a real change can be told apart from ordinary variance.

Why wording must stay fixed

Changing a single word changes which sources are retrieved. A prompt set that drifts produces a trend line that means nothing at all.

The shared measurement

Server logs. Crawler arrival is the one number that tells you something true about both engines at once.

The two opposite mistakes

One supplier will tell you SEO is dead and only AI matters. Another will tell you AI search is a fad worth ignoring. Both are wrong, and in the same way: each has picked a single engine and built a story around it.

Answer engines retrieve from the indexes search optimization builds, so the first claim is structurally impossible. And AI referral traffic, while still around one per cent of the total in the largest study with a stated denominator, is growing while the fixes that capture it cost hours rather than months.

Signs a search optimization company has not caught up — The two opposite errors — dismissing AI search, and declaring SEO dead — are both wrong, and a supplier making either one is not thinking about your site.

SEO is dead — Error. Answer engines retrieve from the indexes SEO builds.

AI search is a fad — Error. Referral share is small but growing and the fixes are cheap.

Buy two suppliers — Error. Duplicate technical work, contradictory reports.

One blended score — Error. Hides which engine is actually working.

Rank tracking proves AI — Error. It cannot see inside an answer.

Publish more articles — Error. Volume is not the lever on the answer engine.

Where the budget should actually go

It depends on two things: how question-shaped your demand is, and how far your particular category has moved toward asking assistants rather than searching. Almost nobody is fully in the second world yet.

For most businesses the honest answer is a majority on the index engine, where the volume still overwhelmingly is, plus the answer-engine fixes that cost hours rather than budget. The split shifts over time; it does not flip.

Where to put the budget — Almost nobody is in the top-right corner yet. Most businesses are somewhere along the diagonal, which is why a single programme covering both beats two specialists.

The shape of the two-engine market — The volume is still overwhelmingly on the index engine. The scarcity, and therefore the leverage, is on the answer engine.

Fits — B2B SaaS. Question-led research, assistant-using buyers.

Fits — Professional services. High-value, question-shaped discovery.

Fits — Healthcare. Question-shaped demand, accuracy critical.

Fits — Manufacturing. Specification queries suit passage retrieval.

Fits — Ecommerce. Index engine dominant, AI shopping emerging.

Fits least — Local trades. Index engine and listings still carry this.

If your demand is keyword-shaped

Most of the budget stays on the index engine. Do the free answer-engine fixes anyway, because they cost hours and remove a real risk.

If your demand is question-shaped

The answer engine deserves real budget, because your buyers are already phrasing their research the way assistants consume it.

If your category has visibly moved

Watch whether competitors appear by name in answers where you do not. That is the signal, and it is checkable monthly for free.

If you are a local trade

The index engine, listings and reviews still carry this, and we would rather say so than sell you a programme that returns less.

Want this done for your site?We run index-engine and answer-engine optimization as one programme, so you never pay twice for the five shared workstreams.

Get a free proposal

What happens to your existing SEO investment

Nothing is wasted. Every pound already spent on indexation, architecture, speed, linking and genuine authority is the substrate the answer engine retrieves from. The AI work sits on top of it rather than replacing it.

What does need revisiting is the habits: publishing to a cadence, chasing keyword density, and treating rank position as the definition of success. Those three do not merely fail to help the answer engine — they actively work against it.

Carries over untouched

Crawl health, indexation, site architecture, page speed, internal linking and topical authority earned over years.

Carries over with a change of emphasis

Content, which shifts from volume toward passages that answer their own headings, and links, which matter for corroboration as much as authority.

Needs retiring

Keyword density targets, publishing cadence as a KPI, and rank position as the sole success measure.

Needs adding

Crawler access verification, scriptless render testing, and a prompt baseline.

What this means for an existing retainer

Usually an adjustment rather than a replacement. Ask your current supplier the two technical questions before assuming you need a new one.

Keyword density targets, publishing cadence as a KPI, and rank position as the sole success measure.

Common objections, answered plainly

Three come up every time: the traffic is too small to matter, it is too early to invest, and nobody can prove any of it works. The first two are questions of timing and the third is a question of method.

On proof specifically: the reason AI visibility claims are so often unfalsifiable is that almost nobody records a baseline before starting. That is a discipline problem, not an inherent property of the channel.

‘AI traffic is only about one per cent’

True, and it is the correct number to use. The argument for acting is not the current share, it is that the highest-impact fixes cost hours and the positions are currently uncontested.

‘It is too early’

For local trades and walk-in retail, we agree and say so. For question-led B2B categories, the buyers are already there.

‘Nobody can prove it works’

They can, if a dated prompt baseline exists with verbatim answers stored. Without one, no claim is falsifiable — which is exactly why we record it in week one.

‘Our rankings are fine’

Rankings and citations are different measurements of different surfaces. A site can hold position one and be absent from every answer a buyer receives.

‘We will wait and see what settles’

Reasonable for strategy, expensive for access. Crawler permissions and rendering should be fixed regardless, because they cost almost nothing and their absence is a silent risk.

Find out where your budget should actually sit

We will check crawl health, AI crawler access and scriptless rendering, take both baselines in the same week, and tell you honestly which engine deserves the money this year.

/ai-visibility-audit

Everything else we have written on search, AI and getting found

Getting found in search

AI, AEO and what is changing

Paid media and lead generation

Websites and design

Choosing and working with an agency

By industry and by situation

Not sure which of these applies to you?Tell us the situation and we will say plainly what we would do first, and what we would not.

Talk it through

The two engines, compared on everything that matters

Search and answer generation compared across the dimensions that change what you should actually do.

One table for the whole distinction
Dimension	Index engine (Google, Bing)	Answer engine (ChatGPT, Perplexity, Copilot, AI Overviews)
What is placed	A page among ten links	A passage inside a written answer
Number of winners	Ten, then more pages	Three or four, no second page
Long tail of positions	Yes, still sends traffic	None. Named or absent
Crawlers	Googlebot, Bingbot	GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, Google-Extended
JavaScript	Usually executed	Frequently not executed
Unit of work	The page and the keyword	The passage and the template
Measurement	Rank tracking, Search Console	Fixed prompt set, identical wording
Paid option	Ads alongside results	None. No inventory exists
Share of demand today	Overwhelming majority	Roughly 1% of web traffic
Competitive density	Very high	Currently low

What overlaps, what does not, and what it means for what you buy

The shared foundation is larger than the differences, which has direct consequences for how a programme should be scoped.

Seven workstreams, and which engine each one serves
Workstream	Index engine	Answer engine	Buy it once or twice?
Indexation and crawl health	Essential	Essential	Once
Site architecture	Essential	Helps	Once
Page speed	Helps	Helps	Once
Internal linking	Essential	Helps	Once
AI crawler access	No effect	Decisive	Answer engine only — and it is free
Scriptless rendering	Minor	Decisive	Answer engine only — and it is the cost
Passage structure	Helps	Decisive	Once, serves both
Entity consistency	Helps	Decisive	Once, serves both
Corroboration	Essential	Essential	Once, serves both
Keyword research	Essential	No effect	Index engine only
Rank tracking	Essential	Cannot measure it	Index engine only
Prompt set	No effect	Essential	Answer engine only

How the generic head terms behave now

What the broad category terms currently return, measured rather than asserted.

What page one actually returns for engine-agnostic phrases
Term	Volume	CPC	What page one now contains
engine optimization service	74,000	$44.66	GEO and AEO service pages at positions 2, 5, 7, 8
search optimization company	22,200	$72.17	Mixed; an AI SEO page ranks at 14
search optimization consultants	8,100	$53.43	An AI SEO page ranks at 8
website optimization services	4,400	$9.09	Traditional SEO services pages
search engine optimisation providers	720	$16.84	A GEO services page at position 3

Two reports, never one score

Why the two engines need separate reporting, and what each report should contain.

What a combined programme reports each month
Stream	Metrics	Source	Why it stays separate
Index engine	Positions, impressions, clicks, conversions	Rank tracker, Search Console	Mature and well understood
Answer engine	Citation frequency, share of voice, sentiment, prompt coverage	Fixed prompt set, stored verbatim	No positions exist to track
Shared	AI and search crawler hits, status codes	Server logs	The one number true of both
Never	A blended visibility score	—	Hides which engine is actually working

Getting found in search

AI, AEO and what is changing

Paid media and lead generation

Websites and design

Choosing and working with an agency

Social, content and brand

By industry and by situation

Frequently asked questions

What is AI search optimization?

Work that makes a site visible in both kinds of engine: the index engines that rank pages among links, and the answer engines that compose a written response and name three or four sources. The technical foundation is shared; crawler access and rendering requirements are not.

Is AI search optimization different from SEO?

It includes SEO and adds two workstreams: allowing and verifying a second set of crawlers, and making pages render without JavaScript. Everything else — indexation, architecture, speed, linking, authority — is shared.

What does a search optimization company do now?

Covers both engines from one technical workstream, and reports them separately: rankings, impressions and clicks for the index engine; citation frequency, share of voice and sentiment for the answer engine.

Should I hire separate SEO and AEO suppliers?

Usually not. Only two of seven workstreams are answer-engine-specific, so two suppliers means paying twice for five of them and receiving two reports that disagree.

Is SEO dead because of AI?

No, and the claim is structurally impossible. Answer engines retrieve from the web indexes SEO builds and maintains, so removing that work removes the foundation the AI work stands on.

Is AI search just a fad?

No. AI referrals are around one per cent of web traffic in the largest study with a stated denominator — small, real, and growing. The fixes that capture it cost hours rather than months, which is what makes it worth doing now.

How much of my budget should go to AI search?

For most businesses, a minority. The volume is still overwhelmingly on the index engine. But the answer-engine fixes are cheap enough that not doing them is a choice rather than a budget constraint.

What are the generic head terms worth?

‘Engine optimization service’ is 74,000 a month at $44.66 per click and ‘search optimization company’ 22,200 at $72.17. Both are engine-agnostic phrases whose page one now includes AI and generative service pages.

Which crawlers does AI search optimization involve?

GPTBot and OAI-SearchBot from OpenAI, ClaudeBot from Anthropic, PerplexityBot, and Google-Extended, which governs Gemini grounding — in addition to Googlebot and Bingbot.

Why does JavaScript matter more for AI search?

Because many AI crawlers do not execute it. A page that assembles its content in the browser returns almost nothing to them, while Google usually renders and indexes it fine.

How is AI search visibility measured?

Against a fixed set of buyer questions re-run with identical wording, with answers stored verbatim. There is no ranked list inside an answer for a rank tracker to track.

Can I use one visibility score for both engines?

You can, and you should not. A blended number is the commonest way weak work in one channel hides behind genuine progress in the other.

What is the cheapest thing to do first?

Check whether your robots.txt blocks the AI crawlers and unblock them if it does. It costs nothing, takes minutes, and it is the difference between being a candidate and being absent.

What is the most expensive thing likely to come up?

Rendering. If your commercial pages only produce content after JavaScript runs, fixing it is an engineering project and it dominates every other line.

Do the same content improvements help both engines?

Passage structure does — writing so each section answers its own heading improves extraction for the answer engine and readability for everyone. Content volume and keyword density do not.

Which businesses should prioritise the answer engine?

B2B SaaS, professional services, healthcare, fintech and manufacturing, where buyers research by asking questions. Local trades still get more from the index engine, listings and reviews.

Can a low-authority site win the generic head terms?

Currently, to a surprising degree. A domain with an authority score of six holds a page-two position on the 74,000-a-month term, because most incumbents have not updated what their service pages cover.

How long before either engine shows movement?

Crawler access shows in logs within days. Rendering and structure take weeks. Rankings and citation frequency both move over months, and citation frequency moves unevenly because model updates are outside anyone’s control.

What should be reported monthly?

Two streams side by side: rankings, impressions and clicks; and citation frequency, share of voice, sentiment and prompt coverage. Plus crawler hits from server logs, which serve both.

Does the answer engine have ads?

No. There is no ad inventory, no submission process and no paid tier. That absence is exactly why the technical work has leverage right now.

What happens to the index engine as AI grows?

It keeps most of the volume for the foreseeable future and remains the retrieval substrate for the answer engine. The split shifts; it does not flip.

What is the single best first step?

Take both baselines in the same week — a rank baseline and a prompt baseline. Neither can be reconstructed favourably later, and without them no claim about progress is falsifiable.

Want this done for your site?We build and maintain the search, content and paid programmes described on this page.

Get a free proposal

Get a free marketing proposal

Tell us what you are trying to grow and we will come back with a plan, not a pitch deck. Same-day reply on weekdays.